Sie sind auf Seite 1von 157

ESCUELA TCNICA SUPERIOR DE INGENIERA (ICAI)

INGENIERO INDUSTRIAL
3D DETAILED FE PARAMETRIC MODEL
FOR A COMPOSITE BOLTED JOINED TEST
UNDER SHEAR-TENSION INTERACTION
Autor: J orge Kraemer vila
Director: M.Sc. B.Eng. Eju Baruah (Airbus Operations GmbH)

Madrid
J unio 2012








ESCUELA TCNICA SUPERIOR DE INGENIERA (ICAI)
INGENIERO INDUSTRIAL
3D DETAILED FE PARAMETRIC MODEL
FOR A COMPOSITE BOLTED JOINED TEST
UNDER SHEAR-TENSION INTERACTION
Autor: J orge Kraemer vila
Director: M.Sc. B.Eng. Eju Baruah (Airbus Operations GmbH)

Madrid
J unio 2012

3D DETAILED FE PARAMETRIC MODEL FOR A COMPOSITE
BOLTED J OINT TEST UNDER SHEAR-TENSION INTERACTION
Autor: Kraemer vila, Jorge
Director: Baruah, Eju
Entidad Colaboradora: Airbus Operations GmbH
Resumen
Hoy en da, el nmero de vuelos est creciendo exponencialmente, por ese motivo es
indispensable encontrar nuevas formas de ahorro de combustible. Los materiales
compuestos avanzados presentes en una era de cambio climtico y rpida disminucin
de los recursos fsiles, no son slo una oportunidad para crear robustas e innovadoras
estructuras que ahorren peso para la industria aeronutica, sino tambin un gran reto de
la ingeniera, como se puede observar en este estudio sobre el comportamiento de Test
ARCAN. La fabricacin, el mantenimiento y la necesidad de estructuras capaces de
soportar carga de manera fiable hacen indispensable el uso de juntas mecnicas. Dada la
falta de experiencia con materiales compuestos en comparacin con el metal en general
y con articulaciones en particular, el diseo de estructuras compuestas fiables es una
cuestin de gran desarrollo as como los costes de experimentacin.
Con el fin de reducir estos costes de experimentacin y por lo tanto reducir el tiempo de
entrega de diseo, avanzados mtodos numricos deben ser establecidos para validar los
resultados de las pruebas con precisin. Con esta finalidad y mediante el uso del cdigo
de elementos finitos Abaqus Explicit, se ha realizado una investigacin sobre el
comportamiento de parmetros fsicos y materiales en una junta mecnica de materiales
compuestos bajo una carga cuasi-esttica a travs el Test ARCAN.

El funcionamiento interno del simulador consiste en una secuencia de comandos
principal, llamada "Terminal", que est a cargo de la gestin de scripts y datos.
Terminal
PartCreator
Retainig
Block
Bolts
Coupon
Nuts
Property
Materials
Sections
Assembly
Meshing
Interaction
Steps
Este script recibe los parmetros de entrada del interfaz de usuario (abajo
mostrado) y gestiona todos los sub-scripts de comandos que envan de vuelta al
usuario la salida solicitada, es decir, el modelo 3D. Tal como se presenta en el
diagrama de flujo anterior, la herramienta se compone de la script de comandos
principal y seis sub-scripts de comandos, cada uno de estos sub-scripts est a
cargo de una de las partes principales del modelo FE generados automticamente
con el pre-procesador Abaqus CAE. Algunos de estos scripts de comandos se
dividen en varias funciones con el fin de armonizar el cdigo.

En una primera fase, el usuario llama al plugin mediante el pre-procesador, la siguiente
interfaz grfica de usuario aparece (arriba mostrado), que permite la seleccin de varios
sub-mens. Dentro de cada sub-men, una amplia gama de modelos y definiciones de
propiedad aparece, que permite la seleccin de las propiedades necesarias para crear el
modelo con las aportaciones correspondientes, y simular el modelo generado.

Para tensin cortante pura, un mtodo simple
utilizando elementos slidos para las capas de
material compuesto unidas por "tie constraint"
ofrece un buen enfoque debido al hecho de que la
rigidez global del modelo est dominado
principalmente por rigidez del laminado, en
particular, dominado por la rigidez en la direccin
1 y 2, este comportamiento queda bien
reproducido por la subrutina material.
Para tensin pura, la aproximacin antes
mencionada es buena para el clculo de la rigidez,
pero ha de ser mejorada con el fin de capturar el
fallo de delaminacin. Por esta razn, la
aproximacin con "cohesive elements" se ha
implementado, con resultados an mejores para la
rigidez y la captura del fallo inter-laminar, el
nico problema que presenta este mtodo es el
enorme costo computacional en comparacin con
la aproximacin anterior. Por esta razn, se ha
implementado la aproximacin de "cohesive
contact" se ha utilizado. Para obtener mejores
resultados con este enfoque, una versin ms
madura del cdigo de elementos finitos Abaqus
Explicit ha de ser probado y utilizado, dnde las
incoherencias numricas durante los anlisis se
han eliminado.
La subrutina material utilizada para este estudio
no tiene en cuenta la diferencia entre el fallo a
tensin y a compresin de la fibra ni de la matriz,
por esta razn se recomienda ejecutar los modelos
utilizando subrutinas de material ms complejas,
ya que el "Enhanced Hashin" o el desarrollado
0 1 2 3 4 5
R
F
1
[
k
N
]
U1[mm]
Solids_wocoh
ExperimentalResults
recientemente "Larc05". La aproximacin con "cohesive elements" parece ser la ms
adecuada para Abaqus 6.10-1, pero presenta un enorme coste computacional. Se puede
utilizar, pero con el fin de minimizar este problema, es necesario establecer una
evolucin tecnolgica-computacional en trminos de mejora de la paralelizacin de los
procesadores, y aumento del nmero CPUs, aunque el aumento del nmero de CPUs
despus de una cierta cantidad no siempre supone una reduccin del coste
computacional. Por esta razn, el uso de Abaqus 6.12-1 es altamente recomendable para
realizar una aproximacin de "cohesive contact" con xito, evitando la incoherencia en
el anlisis numrico entre el comportamiento cohesivo y general de contacto.


3D DETAILED FE PARAMETRIC MODEL FOR A COMPOSITE
BOLTED J OINT TEST UNDER SHEAR-TENSION INTERACTION
Summary
Nowadays the number of flights is growing exponentially because of which it is
indispensable to find new ways of saving fuel. Advanced composite materials present in
an era of climate change and rapidly decreasing fossil resources not only an opportunity
for robust and innovative weight saving structures for the aircraft industry, but also
great challenges in engineering as it can be observed in this study regarding the
behavior of the ARCAN Test. The manufacturing, maintenance and the need of reliable
load carrying structures make indispensable the use of mechanically fastened joints.
Given the lack of experience with composites in comparison to metal in general and
with joints in particular, the design of reliable composite structures is a matter of huge
development as well as experimental costs.
In order to reduce these experimental costs and therefore reduce design laid time,
advanced numerical methods must be established to validate test results accurately.
Based on this assumption and by use of the finite element code Abaqus Explicit, a
numerical and physical parameters investigation on the behavior of a composite bolted
joint under quasi-static loading have been performed through the ARCAN Test.

The internal performance of the simulator consists in a main script, so called
Terminal, which is itself, in charge of script and data management. This script
receives the inputs from the User-Interface Modules and manages all the sub-scripts
sending back to the user the Output requested, that is the 3D model. As presented in
flowchart above, the tool is composed of the main script and six sub-scripts; each of
Terminal
PartCreator
Retainig
Block
Bolts
Coupon
Nuts
Property
Materials
Sections
Assembly
Meshing
Interaction
Steps
these sub-scripts is in charge of one of the main parts of the FE Model automatically
generated with the Abaqus CAE Pre-processor. Some of these scripts are also divided
into several functions in order to harmonize the code.

At a first step, when the plugin-tool is called into the Pre-processor by the end-user, the
following graphical user-interface appears, that allows the selection of several sub-
menus. Under each sub-menu, a range of model and property definitions appears, that
allows the selection of the required properties to create the model with the relevant
inputs, and simulate the generated model.


0 1 2 3 4 5
R
F
1
[
k
N
]
U1[mm]
Solids_wocoh
ExperimentalResults
For pure shear, a simple approach using solid elements for the layers of composite tied
together by tie contact offers a good approach due to the fact that the global stiffness of
the model is dominated mainly by laminate stiffness, in particular dominated by the
stiffness in direction 1 and 2, this behavior is well reproduced by the user subroutine.
For pure tension, the above-mentioned approach
is good for the calculation of the stiffness, but
need to be enhanced with the purpose of
capturing the delamination failure. For this
reason, the approach using cohesive elements has
been implemented, showing even better results
for the stiffness and capturing the inter-laminar
failure, the only problem that this method has is
the enormous computational cost in comparison
with the previous approach. For this reason, the
approach of cohesive contact has been used. To
obtain better results with this approach, a more
matured version of the FE code Abaqus Explicit
is to be tested and used, where such numerical
incoherences during the analyses are said to have
been removed.
The User Material Subroutine used for this study
does not take into account the difference between
tension and compression failure for the fiber
neither for the matrix, for this reason it is
recommended to run the models using more
complex User Subroutine, as the Enhanced
Hashin or the recently developed Larc05.
The approach with cohesive elements seems to be
the most adequate for Abaqus 6.10-1, but presents
enormous computational cost. It can be used but
in order to minimize this problem, it is needed
established a computational evolution in terms
improvement in parallelization of processors and
number of CPU usage, as increasing the number
of CPUs after a certain amount that is not always
laid to reduce computational cost.
For this reason, the use of Abaqus 6.12-1 is
highly recommended to perform a successfully
cohesive contact approach avoiding the
incoherence in the numerical analysis between
cohesive behavior and general contact.
Concluding with the way forward, it would be interesting to implement two different
approaches into the model; one approach for the initial pretension in order to simulate
better the behavior of the joint under and after pretension and another approach for
multi-scale approaches using multi-processors with the aim of distributing wiser the
CPUs, using more for fine meshes and less for coarse meshes.
iv
Agradecimientos
Me gustara dar las gracias a mis padres y amigos. Adems me veo en la obligacin de
dar las gracias a Doa Eju Baruah, que ha supervisado este trabajo, y tantas horas de su
tiempo me ha dedicado. Y me gustara agradecer especialmente a todos los amigos que
me han dado fuerzas en los momentos ms complicados a lo largo de toda mi
trayectoria universitaria, comenzando por aquellos compaeros que me siguieron en mis
inicios en la Universidad Pontificia Comillas y con los que forj un lazo tan especial,
continuando por aquellos que conoc y con los que tanto compart durante mi
experiencia en la Universidad Tcnica de Mnich y por ltimo, pero no menos
importante, aquellos que me han acompaado durante mi experiencia en Airbus
Operations GmbH con los que tantos cafs y frustraciones he compartido. Muchas
gracias a todos.

v


vi
Abstract
The following Thesis aims to study the complex failure phenomenon in composite
bolted joints in shear and tension under quasi-static loading, with the help of Detailed
Finite Element modelling. With this intention a parametric plugin has been programmed
to automate the development of the model. Different modelling methodologies are
investigated in this study, with the emphasis being laid on simulating a pure tension and
a pure shear test for the same joint configuration, and performing comparative analyses
as well as correlations with physical tests. Different numerical approaches have been
implemented and compared in order to obtain the better simulation accuracy-CPU usage
relationship possible.

vii


viii
Table of Contents
1 Introduction 1
1.1 Motivation ...................................................................................................... 1
1.2 Technological Environment ........................................................................... 1
1.3 Project Presentation ....................................................................................... 2
1.3.1 Goals of this Work ................................................................................ 2
1.3.2 Methodology of this Work ................................................................... 3
2 Theoretical Background 5
2.1 Composite Laminate Plate Theory ................................................................. 5
2.1.1 Composite Materials ............................................................................. 5
2.1.1.1 Fiber Materials ............................................................................ 5
2.1.1.2 Matrix Materials .......................................................................... 7
2.1.1.3 Laminate Analysis....................................................................... 9
2.1.2 Micromechanics ................................................................................. 10
2.1.2.1 The Rule of Mixture .................................................................. 10
2.1.2.2 Determination of the Mechanical Properties ............................ 12
2.1.2.2.1 Longitudinal Youngs Modulus E1........................................... 12
2.1.2.2.2 Transversal Youngs Modulus E2............................................. 13
2.1.2.2.3 Transversal Youngs Modulus E3............................................. 14
2.1.2.2.4 Longitudinal Shear Modulus G12 ............................................. 14
2.1.2.2.5 Transversal Shear Modulus G23 ............................................... 16
2.1.2.2.6 Transversal Shear Modulus G13 ............................................... 16
2.1.2.2.7 Longitudinal Poisson Ratio 12 ................................................ 17
2.1.2.2.8 Transversal Poisson Ratio 23 .................................................. 18
2.1.2.2.9 Transversal Poisson Ratio 13 .................................................. 19
2.1.3 Strain-Stress Theory ........................................................................... 19
2.1.3.1 Anisotropic Material ................................................................. 19
2.1.3.2 Monoclinic Material .................................................................. 20
2.1.3.3 Orthotropic Material ................................................................. 21
2.1.3.4 Transversely Isotropic Material ................................................ 22
2.1.3.5 Isotropic Material ...................................................................... 23
2.1.4 Mechanics of a Ply ............................................................................. 24
2.1.4.1 Plane Stress Condition .............................................................. 24
2.1.4.2 Tensor and Matrix Transformation ........................................... 25
2.1.5 Mechanics of Laminates ..................................................................... 26
2.1.5.1 Hypotheses ................................................................................ 26
2.1.5.2 Stiffness Matrices of Thin Laminates ....................................... 28
2.2 Composite Failure Theory ........................................................................... 30
2.2.1 Failure of a Lamina ............................................................................ 30
2.2.1.1 Failure Modes of a Laminate .................................................... 30
ix
2.2.1.1.1 Microbuckling ........................................................................... 30
2.2.1.1.2 Fiber Breakage .......................................................................... 31
2.2.1.1.3 Matrix Cracking ........................................................................ 31
2.2.1.1.4 Delamination ............................................................................. 31
2.2.1.2 Failure Criteria .......................................................................... 31
2.2.2 Maximum Value Failure Theories ...................................................... 32
2.2.2.1 Maximum Stress Theory ........................................................... 32
2.2.2.2 Maximum Strain Theory ........................................................... 32
2.2.3 Interactive Failure Theories ................................................................ 32
2.2.3.1 Tsai-Wu Theory ........................................................................ 32
2.2.3.2 Tsai-Hill Theory ........................................................................ 34
2.2.4 Mechanical Failure Theories .............................................................. 35
2.2.4.1 Hashin-Rotem Criterion ............................................................ 35
2.2.4.2 Hashin Criterion ........................................................................ 36
2.2.5 Delamination Initiation Prediction ..................................................... 36
2.2.6 Damage Progression Models .............................................................. 37
2.2.6.1 Ply-Discounting Approach ........................................................ 38
2.2.6.2 Continuum-Damage-Mechanics Approach ............................... 38
2.3 Composite Bolted Joints in Aircraft Structures ........................................... 38
2.3.1 Mechanically Fastened Joints ............................................................. 38
2.3.1.1 Generalities ............................................................................... 38
2.3.1.2 Failure Modes of Mechanically Fastened Joints ....................... 39
2.3.1.2.1 Tension Failure ......................................................................... 40
2.3.1.2.2 Shear Failure ............................................................................. 40
2.3.1.2.3 Bearing Failure .......................................................................... 41
2.3.1.2.4 Cleavage Failure ....................................................................... 41
2.3.1.2.5 Pull-out Failure ......................................................................... 41
2.3.2 Design Considerations: Material Parameters and Joint Geometry ..... 42
2.3.2.1 Fiber Orientation ....................................................................... 42
2.3.2.2 Stacking Sequence .................................................................... 42
2.3.2.3 Joint Dimensions ....................................................................... 42
2.3.3 Design Considerations: Fastener Parameters ..................................... 42
2.3.3.1 Clamping Torque ...................................................................... 42
2.3.3.2 Friction ...................................................................................... 43
2.3.3.3 Bolt/Hole Clearance .................................................................. 43
2.4 Finite Element Modeling ............................................................................. 43
2.4.1 Finite Element Theory ........................................................................ 44
2.4.1.1 Hypothesis of the Finite Element Analysis ............................... 44
2.4.1.2 Explicit Finite Element Analysis .............................................. 44
2.4.2 Bolted Joints Finite Element Modeling .............................................. 46
2.4.2.1 Pin joints Modeling ................................................................... 46
2.4.2.2 Bolted Joints Modeling ............................................................. 46
x
2.4.3 Interlaminar Damage Modeling ......................................................... 47
2.4.3.1 Fracture Mechanics Theory ...................................................... 47
2.4.3.1.1 Griffith Energy Criterion .......................................................... 47
2.4.3.1.2 Mixed-Mode Delamination ....................................................... 48
2.4.3.2 Virtual Crack Closure Technique ............................................. 49
2.4.3.3 Cohesive Zone Models .............................................................. 50
2.4.3.3.1 Principle .................................................................................... 50
2.4.3.3.2 Constitutive Equations .............................................................. 52
2.4.3.3.3 Mixed-mode Delamination ....................................................... 54
2.4.3.3.4 Comments on the Cohesive Zone Models ................................ 56
3 FE Modeling of the ARCAN Test 57
3.1 Description of the ARCAN Physical Test ................................................ 57
3.1.1 Geometry ............................................................................................ 57
3.1.1.1 Specimen ................................................................................... 57
3.1.1.2 Test Fastener ............................................................................. 58
3.1.1.3 Test Fixture ............................................................................... 59
3.1.1.4 Retaining Blocks ....................................................................... 59
3.1.1.5 Fasteners .................................................................................... 60
3.1.1.6 Semi-Circumferential Parts ....................................................... 60
3.1.2 Test Description .................................................................................. 61
3.1.3 Test Procedure .................................................................................... 61
3.2 Description of the ARCAN FE Virtual Test ............................................. 62
3.2.1 Development of a Tool for a Parametric Detailed FE Model ............ 62
3.2.1.1 Python-Script Flowchart ........................................................... 62
3.2.1.2 User-Interface Modules ............................................................. 63
3.2.2 Geometry ............................................................................................ 65
3.2.2.1 Composite Plate ........................................................................ 65
3.2.2.2 Fastener ..................................................................................... 66
3.2.2.3 Test Fixture ............................................................................... 66
3.2.2.3.1 Retaining Blocks ....................................................................... 66
3.2.2.3.2 Retaining Fasteners ................................................................... 67
3.2.3 Meshing and Element Properties ........................................................ 68
3.2.3.1 Test Fastener ............................................................................. 68
3.2.3.2 Core Vicinity ............................................................................. 69
3.2.3.3 Coupon Core ............................................................................. 70
3.2.3.4 Coupon Outer Cores .................................................................. 70
3.2.3.5 Cohesives (Approach of Cohesive Elements) ........................... 71
3.2.3.6 Retaining Block......................................................................... 72
3.2.3.7 Retaining Fasteners ................................................................... 72
3.2.3.8 Cohesive Elements .................................................................... 73
3.2.4 Constraints & Interactions .................................................................. 75
xi
3.2.4.1 Tie Constraint ............................................................................ 75
3.2.4.2 General Contact ......................................................................... 75
3.2.4.3 Cohesive Behavior .................................................................... 77
3.2.5 User Subroutine .................................................................................. 78
3.2.6 Loading Steps ..................................................................................... 81
3.2.6.1 Initial Pre-tension ...................................................................... 81
3.2.6.2 Final Test Displacement ............................................................ 82
4 Analysis of Experimental Results 83
4.1 Protruding Head Bolt: Pure Tension ............................................................ 83
4.2 Protruding Head Bolt: Pure Shear ................................................................ 84
4.3 Comparison of Tests with Protruding Head Bolt ......................................... 85
4.4 Comparison of Tests with Countersunk Head Bolt ..................................... 87
4.5 Comparison in Terms of Maximum Peak Loads ......................................... 88
5 Analysis of Simulation Results 91
5.1 Solids without Cohesives ............................................................................. 91
5.1.1 Pure Tension ....................................................................................... 91
5.1.2 Pure Shear ........................................................................................... 93
5.2 Cohesive Elements ....................................................................................... 95
5.2.1 Pure Tension ....................................................................................... 95
5.3 Cohesive Contact ......................................................................................... 98
5.3.1 Pure Tension ....................................................................................... 98
5.4 Computational Cost ................................................................................... 100
6 Conclusion and Way Forward 103
6.1 Conclusion ................................................................................................. 103
6.2 Way Forward ............................................................................................. 103
A. Apendices 105
a. Data and Software-code 105
i. Terminal Error! Marcador no definido.
ii. User Material Subroutine [21] 105
b. List of Figures 121
c. List of Tables 125
B. Bibliography 127


xii
Nomenclature

Formelzeichen Unit Kurzbezeichnung
t s Time
v m/s Velocity
N/m
3
Density
S N/m
2
Tensile Strength
E N/m
2
Youngs Modulus
V m
3
Volume
- Specific Volume
- Poisson Ratio
A m
2
Area
m kg Mass
F N Force

i
N/m
2
Tension Stress

i
- Longitudinal Strain
L m Lenght
G
ij
N/m
2
Shear Modulus
N/m
2
Shear Stress
- Shear Strain



xiii


xiv
Abbreviations

FRP Fiber Reinforced Polymers
FE Finite Element
ALLFD Friction Dissipated Energy of the Whole Model
ALLIE Internal Energy of the Whole Model
ALLKE Kinetic Energy of the Whole Model
ALLWK External Energy of the Whole Model
CAE Complete Abaqus Environment
COH3D8 Cohesive 3-Dimensional Element with 8 nodes
CPU Central Processing Unit
C3D8R Continuum 3-Dimensional reduced element with 8 nodes
ETOTAL Total Energy of the Whole Model
FEA Finite Element Analysis
RB Rigid Body
RP Reference Point
SLS Single Lap-Shear
VUMAT User Material for Abaqus Explicit
SDV State Definition Variables
CFRP Composite Fiber Reinforced Plastic
RAM Read Access Memory
xv


1 Introduction 1


1 Introduction
1.1 Motivation
1.2 Technological Environment
A composite is when two or more different materials are combined together to create a
superior and unique material. The first uses of composites date back to 1500 B.C. when
early Egyptians and Mesopotamian settlers used a mixture of mud and straw to create
strong and durable buildings. Straw continued to provide reinforcement to ancient
composite products including pottery and boats.
Later in 1200 A.D., the Mongols invented the first composite bow. Using a combination
of wood, bone, and animal glue, bows were pressed and wrapped with birch bark.
These bows were extremely powerful and highly accurate. Mongolian composite bows
provided Genghis Khan with military dominance, and because of the composite
technology, this weapon was the most powerful weapon on earth until the invention of
gunpowder. But we need to go forward 700 years to the future, to the early 1900s, when
plastics such as vinyl, polystyrene, phenolic and polyester were developed. And in
1935, Owens Corning introduced the first glass fiber, fiberglass. Fiberglass, when
combined with a plastic polymer creates an incredibly strong structure that is also
lightweight. This is the beginning of the Fiber Reinforced Polymers (FRP) industry as
we know it today.
After the First World War, improvements in chemical technology led to an explosion in
new forms of plastics and with it the improvement of the composites, thanks to Brandt
Goldsworthy, often referred to as the grandfather of composites, who developed new
manufacturing processes and products. Goldsworthy invented a manufacturing process
known as pultrusion. Today, products manufactured from this process include ladder
rails, tool handles, pipes, arrow shafts, armor, train floors, medical devices, and more.
As we have seen, the use of composites is not something new, so the question is, why
now, and the easiest answer is because of the weight reduction that composites give us
in the aviation industry. And there are two main reasons to get a weight reduction, to
achieve a reduction of fuel consumption and more importantly, a reduction in
environmental impact.
The Stern Review, 2006, identified that 1.6% of global greenhouse gas emissions come
from aviation and that the demand for air travel will keep rising with our income. To
combat the environmental threat that aviation poses, the Advisory Council for
Aeronautical Research in Europe in 2002 laid out targets to reduce the emission of CO2
(an important greenhouse gas) from an aircraft by 50% by 2020. The reduction of
airframe weight through the extensive use of carbon composites is just one of a range of
technologies that must be deployed to meet such a challenging target. [1]
1 Introduction 2


Tab. 1: Evolution of final energy consumption in transport, by transport mode, 1990-2004, EU-25
(in million toe) [2]

The increased criss-crossing of jet trails we see in the skies reflects the fastest rise in
energy consumption of all transport modes (including maritime transport). As illustrated
in Table 1, between 1990 and 2004 energy consumption rose by 67 % in aviation, this
was considerably greater than the 27 % growth recorded for road transport [2].

1.3 Project Presentation
The structures used in aircrafts to protect the passengers were firstly designed using
metallic materials, the mechanical properties of which and especially the plastic
capabilities are well known. However, since some years now, other materials like
reinforced composite materials have been used in aircraft key structure like the fuselage,
and especially in the new generation airplane. Because of this, it is necessary to address
the failure behavior of composite material structures.
1.3.1 Goals of this Work
The global idea in this research project consists in the study of composite bolted joints
failure, in order to be able to reproduce the test results with detailed FE models, to late
perform sensitivity studies. To try to get the better understanding of the failure
mechanisms in the fuselage, a detailed modeling of a composite bolted joint will be
performed using the ARCAN test.
This study aims at:
Developing a friendly non-technical user oriented interface.
Developing a 3D parametric simulator.
Reproducing and capturing the behavior of the fastened joints.
Validation of the experimental results.
1 Introduction 3


1.3.2 Methodology of this Work
In order to develop this parametric 3D ARCAN test simulator, a highly operative
scripting model was built up, using the following flow charts.

Tab. 2: Functional flowchart between the user and the tool.

This model consists in dividing each task of the script in separated sub-scripts with less
than 200 code-lines obtaining the following advantages:
Reduction in computational cost.
Increment in the accuracy in scripting failure detection.
User-friendly further optimization of the tool.
Easy-to-use scripting interfaces and simulator for a new end-user.
In the following sections, the tests and the finite element models will be presented and
correlation of the FE results with the test results will be performed.


User Terminal
1 Introduction 4




2 Theoretical Background 5


2 Theoretical Background
2.1 Composite Laminate Plate Theory
2.1.1 Composite Materials
Generally speaking, a composite material signifies that two or more components with
different properties are combined on a macroscopic scale to form a useful third material with
properties that are not attainable with the individual components, as has been used by man for
thousands of years. The advantage of composite materials is that, they usually exhibit the best
qualities of their components and often some qualities that neither component possesses. Not
all these material properties are improved at the same time. In fact, some of the properties are
in conflict with another.
2.1.1.1 Fiber Materials
Polymers have low stiffness and are ductile, in contrast, ceramics and glasses are stiff and
strong but are catastrophically brittle. In fibrous composites, we exploit the great strength of
ceramics while avoiding the brittle failure of fibers, leading to a progressive, not a sudden,
failure. If the fibers of a composite are aligned along the loading direction, the stiffness and
the strength are roughly speaking, an average of those of the matrix and fibers, weighted by
their volume fractions. But not all composite properties are just a linear combination of these
of the components. Their great attraction lies in the fact that, frequently, something extra is
gained.
Tab. 3: Fiber and Wire Properties [3]

The fibers are subjected to special finished surface treatments to prevent fiber damage under
contact with processing equipment, to provide surface wetting when fibers are combined with
matrix materials and to improve the interface bond between fibers and matrices.
2 Theoretical Background 6


The most commonly encountered surface treatments are chemical sizing performed during the
basic fiber formation operation and resulting in a thin layer applied to the surface of the fiber,
surface etching by acid, plasma or corona discharge, and coating of fiber surface with thin
metal or ceramic layers. With only a few exceptions (e.g., metal fibers), individual fibers,
being very thin and sensitive to damage, are not used in composite manufacturing directly, but
in the form of tows (rovings), yarns and fabrics. A unidirectional tow (roving) is a loose
assemblage of parallel fibers consisting usually of thousands of elementary fibers. Two main
designations are used to indicate the size of the tow, namely the K-number that gives the
number of fibers in the tow (e.g., 3K tow contains 3000 fibers) and the tex-number which is
the mass in grams of 1000 m of the tow. The tow tex-number depends not only on the number
of fibers but also on the fiber diameter and density. A yarn is a fine tow (usually it includes
hundreds of fibers) slightly twisted (about 40 turns per meter) to provide the integrity of its
structure necessary for textile processing. Yarn size is indicated in tex-numbers or in textile
denier-numbers (den) such that 1tcx = 9Jcn. Continuous yarns are used to make fabrics with
various weave patterns. [4]

Fig. 2.1: Stress-strain curves for typical fibers used in reinforced composite
The processability is an important characteristic of fibers and can be evaluated as the ratio,
K
p
=
c
s
c
of the strength demonstrated by fibers in the composite structure, to the strength of
fibers before they were processed. This ratio depends on the fibers ultimate elongation,
sensitivity to damage, and manufacturing equipment causing the damage of fibers. Boron and
high-modulus carbon fibers are the most sensitive to operational damage, possessing
relatively low ultimate elongation 5.

2 Theoretical Background 7



Fig. 2.2: Fiber processability tests [5]: (a) straight tow, (b) tow with a loop and (c) tow with a knot
To evaluate fiber processability under real manufacturing conditions, three simple tests are
used -tension of a straight dry tow, tension of tows with loops, and tension of a tow with a
knot (see Figure 2.2).
2.1.1.2 Matrix Materials
To utilize high strength and stiffness of fibers in a monolithic composite material suitable for
engineering applications, fibers are bound with a matrix material whose strength and stiffness
are naturally, much lower than those of fibers. Matrix materials provide the final shape of the
composite structure and govern the parameters of the manufacturing process. Optimal
combination of fiber and matrix properties should satisfy a set of operational and
manufacturing requirements that sometimes are of a contradictory nature and have not been
completely met yet in existing composites. The most commonly used matrices are carbon,
ceramic, glass, metal, and polymeric. Each has special appeal and usefulness, as well as
limitations. Richardson [6] presents a comprehensive discussion of matrices, which guided
the following presentation.
2 Theoretical Background 8


Tab. 4: Types of Matrix

Carbon Matrix. A carbon matrix has a high heat capacity per unit weight.
Ceramic Matrix. A ceramic matrix is usually brittle.
Glass Matrix. Glass and glass-ceramic composites usually have an elastic modulus
much lower than that of the reinforcement.
Metal Matrix. A metal matrix is especially good for high-temperature use in
oxidizing environments. There are three classes of metal matrix composites:
o Class I. The reinforcement and matrix are insoluble.
o Class II. The reinforcement and matrix exhibit some solubility and the
interaction will alter the physical properties of the composite.
o Class III. The most critical situations in terms of matrix and reinforcement are
in this class.
Polymer Matrix. Polymeric matrices are the most common and least expensive. They
are found in nature as amber, pitch, and resin. They are a low-density material.
Because of low processing temperatures, many organic reinforcements can be used.
There are two classes of polymer matrix composites:
o Thermoplastic. A thermoplastic can be remolded to a new shape when
it is heated to approximately the same temperature at which it was
formed.
o Thermoset. A thermoset cannot be remolded after it has been
processed.

Matrix
Carbon
Ceramic
Glass
Metal
ClassI
ClassII
ClassIII
Polymer
Thermoplastic
Thermoset
2 Theoretical Background 9


2.1.1.3 Laminate Analysis
As it is shown in Figure 2.3, the analysis of a laminate is a
multi-scale process, starting from the raw materials, the
matrix and the fibers, elements from which the lamina is
built. It is possible to distinguish this category in three
divisions, depending on the allocation of the fibers: lamina
with unidirectional fibers, lamina with woven fibers
(biaxial weave) and lamina with woven fibers (triaxial
weave), as shown in Figure 2.4. To determinate the
properties of the lamina, the use of micromechanics is
needed. It relies on the study of the interaction of the
constituent materials based on models that uses simple
assumptions regarding the lamina and/or experimental
methods.




Fig. 2.3: The levels of analysis for a structure made of laminated composite. [7]

Fig. 2.4: Composite material systems. [7]
When the properties of the lamina are known, the analysis continues with the laminate, which
is built by stacking several laminas with different orientations. The properties of the laminate
are calculated through macromechanics using the mechanical properties of the laminas. The
composite material is assumed homogeneous, therefore the contribution of the constituent
materials are detected as averaged properties of the composite material.

2 Theoretical Background 10


2.1.2 Micromechanics

Fig. 2.5: Illustration of the matrix, fiber, and void volumes. [7]
Micromechanical approach is used to estimate the mechanical properties of composite
materials from the known values of the properties of the fiber and the matrix. There are three
major categories of micromechanical approaches: (a) mechanics of material models based on
simplifying assumptions that make it unnecessary to specify in detail the stressstrain
distributions, (b) elasticity models requiring that the stresses and strains be determined at the
micromechanical level, and (c) empirical expressions resulting from curve-fitting elasticity
solutions or data.
2.1.2.1 The Rule of Mixture

We consider the volume V of an element shown in Figure 2.5. The volume of this element is

I = I
]
+I
m
+I

(2. 1)

where the subscripts f, m and v refer to the fiber, the matrix, and the void. It is convenient to
introduce the volume fractions as follows:

:
]
=
v
]
v
:
m
=
v
m
v
:

=
v

v
(2. 2)

With the equations A.1 and A.2, it can be deduced:

:
]
+:
m
+:

= 1 (2. 3)

When the void fraction is negligible (:

= u), we have
2 Theoretical Background 11



1 = :
]
+:
m
(2. 4)

By taking in consideration the characteristics of the element 1, the following relationships can
be deduced, using the volume fraction and A, A
f
, A
m
, respectively the composite, the fiber
and the matrix cross-sectional areas:

A = I
2
= A
]
+A
m
(2. 5)

:
]
=
A
]
A
:
m
=
A
m
A
(2. 6)

The mass of the element, by neglecting the void mass in Figure 2.6 is

m = m
]
+m
m
= p
]
I
]
+p
m
I
m
(2. 7)

where
f
and
m
are the fiber and the matrix densities, respectively. The density of the
composite is

p
comp
=
m
v
= p
]
:
]
+p
m
:
m
(2. 8)

Fig. 2.6: Representative elements: Element 1 (left) and Element 2 (right). [7]
2 Theoretical Background 12


In the following, as is taken approach to consider unidirectional, fiber-reinforced composites
without voids. Deriving the properties using two types of elements, as shown in Figure 2.6,
element 1 contains a single fiber bundle of circular cross section, element 2 consists of a fiber
layer sandwiched between two layers of matrix material. In element 2, the fiber and matrix
volumes are the same as in element 1.
2.1.2.2 Determination of the Mechanical Properties
2.1.2.2.1 Longitudinal Youngs Modulus E1
Element 1 is subjected to a force F1 in the fiber direction. The force, distributed over the
surface, is (Fig.); with
1
the normal stress averaged on the cross-section with area A, we can
decompose the force into two parts, one carried out for the matrix and the other one for the
fiber:

F
1
=
1
A =
I
A
I
+
m
A
m
(2. 9)

When the Poisson effect is neglected, the normal stresses in the composite
1
, in the fiber
bundle
f1
, and in the matrix
m1
are:

o
1
= e
1
E
1
o
]1
= e
]1
E
]1
o
m1
= e
m1
E
m
(2. 10)

where E
1
and E
f1
are the composite and the fiber longitudinal Youngs modulus and E
m
is the
matrix Youngs modulus respectively. Using the hypothesis of identical deformation in the
longitudinal direction (e
1
= e
]1
= e
m1
) Equations (2.9) and (2.10) give:

E
1
=
A
]
A
E
]1
+
A
m
A
E
m
(2. 11)

Finally, using the Equation (2.6), the rule of mixture for the longitudinal Youngs modulus,
depending on the fiber and matrix modulus, is obtained:

E
1
= :
]
E
]1
+:
m
E
m
(2. 12)



2 Theoretical Background 13


2.1.2.2.2 Transversal Youngs Modulus E2
In Element 2, the inner layer is a sheet of fiber with the same volume as the fiber bundle in
Element 1. The length and the transverse Youngs modulus of this inner layer in x
2
direction
are L
f
and E
f2
respectively, and the length and the Youngs modulus of the outer layers are
I
m
=
(L-L
]
)
2
and E
m
, and the lateral surface of the element is A = I
2
.

Fig. 2.7: Element 1 subjected to a force in the fiber direction (left), and Element 2 subjected to a force in
the transverse direction (right). [7]
The normal stress in the transverse direction is:

o
2
=
P
2
A
= e
2
E
2
=
L
L
E
2
(2. 13)

where E
2
is the transverse Youngs modulus of the element and
2
is the average transverse
normal strain. The change in the length of the element is

I = 2 I
m
e
m2
+I
]
e
]2
(2. 14)

When the Poisson effect is neglected, the transverse normal strains in the composite
2
, in the
fiber bundle
f2
, and in the matrix
m2
are:

e
]2
=
c
]2
L
]2
e
m2
=
c
m2
L
m
(2. 15)

By introducing equations (2.14) and (2.15) into equation (2.13), we obtain:

e
2
=
2L
m
c
m2
LL
m
+
L
]
c
]2
LL
]2
(2. 16)

2 Theoretical Background 14


The transverse normal stresses in the composite
2
, the matrix
m2
, and the fiber
f2
layers are
equal as follows:

o
2
= o
]2
= o
m2
(2. 17)

The matrix and fiber volume fractions are:

:
]
=
L
]
L
:
m
=
L
m
L
(2. 18)

By substituting equations (2.16) and (2.17) into equation (2.16), we obtain the transverse
Youngs modulus

E
2
= _

]
L
]2
+

m
L
m
]
-1
(2. 19)

2.1.2.2.3 Transversal Youngs Modulus E3
By using the same theoretical procedure as for the longitudinal Youngs modulus but using
the representative element 2. , we obtain:

E
3
= :
]
E
]3
+:
m
E
m
(2. 20)

2.1.2.2.4 Longitudinal Shear Modulus G12
We consider Element 2. The length and the longitudinal shear modulus of the inner layer are
L
f
and G
f12
respectively, and the length and the shear modulus of the outer layers are I
m
=
(L-L
]
)
2
and G
m
. The element is subjected to a shear force F
12
(see Figure 2.8) distributed
uniformly across the surface A = I
2
.
2 Theoretical Background 15



Fig. 2.8: Element 2 subjected to a shear force (left) and deformation of the top (ijkl) surface (right). [7]
The shear stress
m
=
P
12
A
is:

12
=
P
12
A
= y
12
0
12
(2. 21)

where G
12
is the longitudinal shear modulus of the element and
12
is the average shear strain,

y
12
tany
12
=
L
L
(2. 22)

The change in the length of the element due to the shear deformations of the matrix and fiber
layers is:

I = 2 I
m
y
m12
+I
]
y
]12
(2. 23)

The shear strains in the composite
2
, in the fiber bundle
f2
, and in the matrix
m2
are:

y
]12
=
:
]12
u
]12
y
m12
=
:
m12
u
m12
(2. 24)

The surface where the shear stresses in the composite are applied, the matrix and the fibers are
the same, thus:

12
=
]12
=
m12
(2. 25)

2 Theoretical Background 16


By introducing equations (2.22), (2.23), (2.24) and (2.25) into equation (2.21), we obtain the
transverse Youngs modulus:

0
12
= _

]
u
]12
+

m
u
m
]
-1
(2. 26)

2.1.2.2.5 Transversal Shear Modulus G23
By using the same theoretical procedure as for the longitudinal Shear modulus but using the
representative element 2, we obtain:

0
23
= _

]
u
]23
+

m
u
m
]
-1
(2. 27)

2.1.2.2.6 Transversal Shear Modulus G13
By using the same reasoning as for the longitudinal Youngs modulus but using the
representative element 2. Taking into account that the shear deformations are indeed
equal (y
12
= y
]12
= y
m12
), the Shear force is distributed on the top surface, thus:

13
I =
]13
I
]
+
m13
I
m
(2. 28)

where
m12
and
f12
are the shear strains in the matrix and fiber layers as follows:

]13
= y
]13
0
]13

m13
= y
m13
0
m
(2. 29)

It can be deduced from equations 2.28 and 2.29 the transverse Shear modulus G
23
:

0
13
= :
]
0
]13
+:
m
0
m
(2. 30)

2 Theoretical Background 17


2.1.2.2.7 Longitudinal Poisson Ratio 12

Fig. 2.9: Deformation of Element 1 subjected to an axial force in the x
1
direction. [7]

As shown in Figure 2.9, element 1 is subjected to a force F1 in the longitudinal x1 direction.
Because of this force the element deforms, as illustrated in Figure 2.9. The sides of the
deformed element are (L +L
1
), (L -L
2
) and (L -L
3
). The change in the cross-sectional
area A of the face ilkj is:

A = A -(I -I
2
) (I -I
3
) = A
]
+A
m
(2. 31)

where A
I
and A
m
are respectively the change in the cross-sectional areas of the fiber and
the matrix. By definition, the normal strains in the x
2
and x
3
directions are:

e
3
= -
L
2
L
= -v
12
e
1
e
3
= -
L
3
L
= -v
13
e
1
(2. 32)

By symmetry of element 1, it is assumable that I
2
= I
3
and therefore:

e
2
= e
3
v
12
= v
13
(2. 33)

By introducing equations 2.32 and 2.33 in 2.31, the change in the cross-sectional area A
becomes:

A = A (e
2
+e
3
) = 2 A e
2
(2. 34)

2 Theoretical Background 18


By the same reasoning, change in the cross-sectional area of the fiber and the matrix are:

I
m
= 2 A
m
e
m2
I
]
= 2 A
]
e
]2
(2. 35)

By combining equations 2.31, 2.32 and 2.33, it is obtained that:

v
12
= -
s
2
s
1
= -
A
2As
1
(2. 36)

By combining equations 2.35 and 2.36, it is obtained that:

v
12
= -
A
]
s
]2
As
1
-
A
m
s
m2
As
1
(2. 37)

Taking into account that the deformation in the composite, the fiber and the matrix are the
same in direction 1:

e
1
= e
]1
= e
m1
(2. 38)

Using the same procedure as for 2.36, the longitudinal Poisson ratio for the fiber and for the
matrix can be expressed as:

v
]12
= -
s
]2
s
]1
v
m
= -
s
m2
s
m1
(2. 39)

Finally, using the equations 2.37, 2.38 and 2.39, the expression of the longitudinal Poisson
ratio is:

v
12
= u
m
v
m
+u
]
v
]12
(2. 40)

2.1.2.2.8 Transversal Poisson Ratio 23
For a transversely isotropic material, the transverse shear and the Youngs modulus are
related by the following expression:
2 Theoretical Background 19



0
23
=
L
2
2(1+v
23
)
(2. 41)

For the element 1, which is considered isotropic in the plane x
2
, x
3
, it can be deduced:

v
23
=
L
2
2u
23
-1 (2. 42)

2.1.2.2.9 Transversal Poisson Ratio 13
By using the same theoretical procedure as for the longitudinal Youngs modulus but using
the representative element 2. We obtain:

v
13
= v
12
= u
m
v
m
+u
]
v
]12
(2. 43)

2.1.3 Strain-Stress Theory
In a composite material, the fibers may be oriented in an arbitrary manner. Depending on the
arrangement of the fibers, the material may behave differently in different directions.
According to their behavior, composites may be characterized as generally anisotropic,
monoclinic, orthotropic, transversely isotropic, or isotropic. In the following, we present the
stressstrain relationships for these types of materials under linearly elastic conditions.
2.1.3.1 Anisotropic Material
A material is referred to as generally anisotropic, when there are no
symmetry planes with respect to the alignment of the fibers in the material.
For a linear elastic material, the following generalized Hookes law
expressed in stiffness.



Fig. 2.10: Anisotropic fiber orientation for plane-strain condition.
In the local coordinate system it can be used in the coordinate system (x
1
, x
2
, x
3
), the stress-
strain relationships are:
2 Theoretical Background 20


`
1
1
1
1

23

13

12
1
1
1
1
1
=
l
l
l
l
l
l
Q
11
Q
12
Q
13
Q
14
Q
15
Q
16
Q
21
Q
22
Q
23
Q
24
Q
25
Q
26
Q
31
Q
32
Q
33
Q
34
Q
35
Q
36
Q
41
Q
42
Q
43
Q
44
Q
45
Q
46
Q
51
Q
52
Q
53
Q
54
Q
55
Q
56
Q
61
Q
62
Q
63
Q
64
Q
65
Q
66
1
1
1
1
1
1

`
1
1
1
1

1

23

13

12
1
1
1
1
1
(2. 44)


where C
ij
are the elements of the stiffness matrix |] in the x
1
, x
2
, x
3
local coordinate system.

`
1
1
1
1
e
1
e
2
e
3
y
23
y
13
y
12
1
1
1
1
1
=
l
l
l
l
l
l
S
11
S
12
S
13
S
14
S
15
S
16
S
21
S
22
S
23
S
24
S
25
S
26
S
31
S
32
S
33
S
34
S
35
S
36
S
41
S
42
S
43
S
44
S
45
S
46
S
51
S
52
S
53
S
54
S
55
S
56
S
61
S
62
S
63
S
64
S
65
S
66
1
1
1
1
1
1

`
1
1
1
1
o
1
o
2
o
3

23

13

12
1
1
1
1
1
(2. 45)

There is of course a relationship between the stiffness and the compliance matrices, indeed:

|S] = |]
-1
(2. 46)

It can be shown that the matrices are symmetrical, thus:

]
=
]
and S
]
= S
]
i = 1, , n (2. 47)

These relationships reduce the number of independent parameters from 36 to 21.
2.1.3.2 Monoclinic Material
A material is referred to as generally monoclinic, when a plane of
symmetry exists. This means that when the material is subjected to a
normal stress in its plane of symmetry, the shear strain that could occur in
the material on one side of the plane of symmetry is compensated by the
response on the other side by the stress and is thus equal to zero.


Fig. 2.11: Monoclinic fiber orientation for plane-strain condition.
2 Theoretical Background 21



|S] =
l
l
l
l
l
l
S
11
S
12
S
13
u u S
16
S
21
S
22
S
23
u u S
26
S
31
S
32
S
33
u u S
36
u u u S
44
S
45
u
u u u S
54
S
55
u
S
61
S
62
S
63
u u S
66
1
1
1
1
1
1
(2. 48)

Using the Youngs and Shear modulus and Poisson ratios, it can be drawn:

|S] =
l
l
l
l
l
l
l
l
l
l
1
E
1
-
u
21
E
2
-
u
31
E
3
u u
u
61
G
12
-
u
12
E
1
1
E
2
-
u
32
E
3
u u
u
62
G
12
-
u
13
E
1
-
u
23
E
2
1
E
3
u u
u
63
G
12
u u u
1
G
23
u
S4
G
13
u
u u u
u
4S
G
23
1
G
13
u
u
16
E
1
u
26
E
2
u
36
E
3
u u
1
G
12
1
1
1
1
1
1
1
1
1
1
(2. 49)

The number of independent terms is now reduced to 13.
2.1.3.3 Orthotropic Material
A material is referred to as generally orthotropic, when having
three planes of symmetry with respect to the alignment of the
fibers; two-by-two are perpendicular. This configuration enables
new simplifications in |S] and |]. In addition to the previous
terms equal to zero, the shear strain implied by normal stresses in
the remaining planes of symmetry is also equal to zero.

Fig. 2.12: Orthotropic fiber orientation for plane-strain condition.

|S] =
l
l
l
l
l
l
S
11
S
12
S
13
u u u
S
21
S
22
S
23
u u u
S
31
S
32
S
33
u u u
u u u S
44
u u
u u u u S
55
u
u u u u u S
66
1
1
1
1
1
1
(2. 50)

Using the Youngs and Shear modulus and Poisson ratios, it can be drawn:
2 Theoretical Background 22



|S] =
l
l
l
l
l
l
l
l
l
l
1
L
1
-
u
21
L
2
-
u
31
L
3
u u u
-
u
12
L
1
1
L
2
-
u
32
L
3
u u u
-
u
13
L
1
-
u
23
L
2
1
L
3
u u u
u u u
1
u
23
u u
u u u u
1
u
13
u
u u u u u
1
u
12
1
1
1
1
1
1
1
1
1
1
(2. 51)

The number of independent terms is now reduced to 9.
2.1.3.4 Transversely Isotropic Material
A transversely isotropic material has three planes of symmetry and, as such, it
is orthotropic. In one of the planes of symmetry the material is treated as
isotropic. In this case, perpendicular-to-the-fibers planes can be treated as
orthogonal. The compliance matrix is simplified as follows:


Fig. 2.13: Transversely Isotropic fiber orientation for plane-strain condition.

|S] =
l
l
l
l
l
l
S
11
S
12
S
13
u u u
S
21
S
22
S
23
u u u
S
31
S
32
S
33
u u u
u u u 2(S
22
-S
23
) u u
u u u u S
66
u
u u u u u S
66
1
1
1
1
1
1
(2. 52)

Using the Youngs and Shear modulus and Poisson ratios, it can be drawn:

|S] =
l
l
l
l
l
l
l
l
l
l
l
1
L
1
-
u
21
L
2
-
u
31
L
3
u u u
-
u
12
L
1
1
L
2
-
u
32
L
3
u u u
-
u
13
L
1
-
u
23
L
2
1
L
3
u u u
u u u
2(1+u
23
)
L
2
u u
u u u u
1
u
13
u
u u u u u
1
u
12
1
1
1
1
1
1
1
1
1
1
1
(2. 53)

2 Theoretical Background 23


The following equation can also be deduced:

|Q] =
l
l
l
l
l
l
l
Q
11
Q
12
Q
13
u u u
Q
21
Q
22
Q
23
u u u
Q
31
Q
32
Q
33
u u u
u u u
(Q
22
-Q
23
)
2
u u
u u u u Q
66
u
u u u u u Q
66
1
1
1
1
1
1
1
(2. 54)

The number of independent terms is now reduced to 5.

2.1.3.5 Isotropic Material
The isotropic material has no preferred directions, so every plane is a plane of
symmetry. When a composite is built up of many plies equally oriented, this
behavior can be assumed. For an isotropic material, the coordinate system chosen
is indifferent and |S] and |] present their simplest form:


Fig. 2.14: Isotropic fiber orientation for plane-strain condition.

|S] =
l
l
l
l
l
l
S
11
S
12
S
13
u u u
S
21
S
22
S
23
u u u
S
31
S
32
S
33
u u u
u u u 2(S
22
-S
23
) u u
u u u u 2(S
22
-S
23
) u
u u u u u 2(S
22
-S
23
)1
1
1
1
1
1
(2. 55)

what can be simplified as:

|S] =
l
l
l
l
l
l
l
l
l
l
1
E
-
u
E
-
u
E
u u u
-
u
E
1
E
-
u
E
u u u
-
u
E
-
u
E
1
E
u u u
u u u
2(1+u)
E
u u
u u u u
2(1+u)
E
u
u u u u u
2(1+u)
E
1
1
1
1
1
1
1
1
1
1
(2. 56)

2 Theoretical Background 24


In this case, there are still 12 non-zero terms, but only 2 independents, and . And also:

|Q] =
l
l
l
l
l
l
l
l
Q
11
Q
12
Q
13
u u u
Q
21
Q
22
Q
23
u u u
Q
31
Q
32
Q
33
u u u
u u u
(Q
22
-Q
23
)
2
u u
u u u u
(Q
22
-Q
23
)
2
u
u u u u u
(Q
22
-Q
23
)
2
1
1
1
1
1
1
1
1
(2. 57)

2.1.4 Mechanics of a Ply
Ply or lamina is the simplest element of a composite material, an elementary layer of
unidirectional fibers in a matrix formed while a unidirectional tape impregnated with resin is
placed on the surface of the tool providing the shape of a composite part.
2.1.4.1 Plane Stress Condition
A ply can be considered as a thin plate, approximating the stresses in the lamina with a plain-
stress state. It is convenient to work in the local coordinate system, to take advantage from the
possible planes of symmetry. In case of plain-stress condition, the normal-to-the-plane stress
and the out-of-the-plane stresses are equal to zero:

o
3
= u
13
= u
23
= u (2. 58)

Therefore, by using the equation A.44:

_
o
1
o
2

12
_ = _

11

12

16

12

22

26

16

26

66
_ _
e
1
e
2
y
12
_ (2. 59)

The same kind of relationship with compliance matrix can be drawn:

_
e
1
e
2
y
12
_ = _
S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_ _
o
1
o
2

12
_ (2. 60)

Still with:

2 Theoretical Background 25


|] = |S]
-1
(2. 61)

Then, being in the local coordinate system, according to the symmetries of the material, the
matrices can be simplified referring to the expressions given in section 2.1.3. The stress and
strain tensors, and respectively, have been expressed in the local coordinate system. The
next step is to transform these matrices and tensors in the global coordinate system.
2.1.4.2 Tensor and Matrix Transformation
For the purpose of getting the expression of [Q] and [S] expressed in the Cartesian coordinate
system (x, y, z), base transformation matrices are used.

Fig. 2.15: Coordinate Base Transformation
For in-plane stress condition, this transformation is simple because the two coordinate
systems have usually the same normal and a rotation of . By noting with [T

] and [T

] the
transformation matrices for stresses and strains respectively from the stress and strain tensors
expressed in the Cartesian coordinate system and to and , it can be expressed in the
form:

o = |I
c
] o s = |I
s
] s (2. 62)

Where the transformation matrices are:

|I
c
] = _
c
2
s
2
2cs
s
2
c
2
-2cs
cs -cs c
2
-s
2
_ |I
s
] = _
c
2
s
2
cs
s
2
c
2
-cs
2cs -2cs c
2
-s
2
_ (2. 63)

with the notation:

2 Theoretical Background 26


c = cos 0 s = sin0 (2. 64)

Using the matrices of the equation 2.63, the expression of [Q] and [S] in the Cartesian
coordinate system (x, y, z) can be deduced , where they will respectively be noted
as |

] and |S

]:

|

] = _

11

12

16

12

22

26

16

26

66
_ = |I
c
]|]|I
s
]
-1
(2. 65)

|S

] = |I
s
]|]|I
c
]
-1
(2. 66)

2.1.5 Mechanics of Laminates
The behavior of a laminate is characterized by three stiffness matrices [A], [B], and [D]. The
expression of these matrices will be determined in the following pages by considering flat
laminates undergoing small deformations. The hypotheses used are hereafter presented.
2.1.5.1 Hypotheses
The analysis will be made here in the frame of flat laminates subjected to small deformations
and relies on the laminate plate theory. Then, the following assumptions can be made [8]:
The strain varies linearly across the laminate.
The out-of-plane shear deformations are negligible.
The out-of -plane stress
z
and the shear stresses
xz
and
yz
are small compared to the
in-plane normal
x
,
y
and the shear stress
xy
.
By using these assumptions, it is possible to consider plane-stress conditions. Thus
relationships which were determined in section 2.1.4 can be used here. Similarly to the
classical plate theory, a referred plane is considered, that corresponds to the plane (xy). Then
it is convenient to take the mid-plane of the laminate, shown in Figure 2.16 (left). In this plane
the stresses are:

x
o
=
0u
o
0x

y
o
=
0v
o
0y

xy
o
=
0u
o
0y
+
0v
o
0x
(2. 67)
Where u and v are respectively the displacements along the x and y directions and the
superscript refers to the reference plane.
The Kirchhoff hypothesis, according to which normals to the reference plane remain straight
under deformations, is here used. This is illustrated in Figure 2.16. The angle of rotation of
the normal to the reference plane
xz
and
yz
can be determined:
2 Theoretical Background 27


xz
=
0w
o
0x

yz
=
0w
o
0y
(2. 68)

where w designates the out-of-plane displacement of the reference plane. The total
displacement of the reference plane along the x and y directions can be written as follows:

u = u
o
- z
xz
= u
o
-z
0w
o
0x
v = v
o
-z
yz
= v
o
-z
0w
o
0y
(2. 69)

The strains of the laminate are defined as:

x
=
0u
0x

y
=
0v
0y

xy
=
0u
0y
+
0v
0x
(2.70)

By introducing the equation 2.69 into the equation 2.70, it is obtained:

x
=
0u
o
0x
-z
0
2
w
o
0x
2
(2. 71)

y
=
0v
o
0y
-z
0
2
w
o
0y
2
(2. 72)

xy
=
0u
o
0x
+
0v
o
0y
-2z
0
2
w
o
0x0y
(2. 73)

Thus, the expression in a matrix form can be written as follows:

_
e
x
e

y
x
_ = _
e
x
o
e

o
y
x
o
_ +z _

x
_ (2. 74)
Where
x
,

and
x
are the curvature of the reference plane of the plate.
2 Theoretical Background 28



Fig. 2.16: Laminate under plane-stress conditions (left) and deformation of the plane in the x-z plane
(right).

2.1.5.2 Stiffness Matrices of Thin Laminates
The in-plane forces N and moments M acting on a very small element of the laminate are
given by the following relationships:

N
x
= ]
x
Jz
h
t
-h
b
N

= ] o

Jz
h
t
-h
b
N
x
= ]
x
Jz
h
t
-h
b
(2. 75)

H
x
= ] z
x
Jz
h
t
-h
b
H

= ] zo

Jz
h
t
-h
b
H
x
= ] z
x
Jz
h
t
-h
b
(2. 76)

Where the terms h
b
and h
t
refer to the distances from the reference plane to the surfaces of the
plate, as is shown in Figure. Furthermore, the following relation in (x, y, z) for a plane is
already known:

_
o
x
o

x
_ = |

] _
e
x
e

y
x
_ (2. 77)

With |

], the stiffness matrix in the Cartesian coordinate system (x, y, z), given in Equation
2.65, where is expressed as a function of [Q] according to the section 2.1.5. By substituting
the equations 2.74 and 2.78 into 2.75 and 2.76, it is obtained:

_
N
x
N

N
x
_ = ] |

]Jz _
e
x
o
e

o
y
x
o
_ +] z|

]Jz _

x
_
z z
(2. 78)
2 Theoretical Background 29



_
H
x
H

H
x
_ = ] z|

]Jz _
e
x
o
e

o
y
x
o
_
z
+] z
2
|

]Jz _

x
_
z
(2. 79)

We can simplify the previous equations using the stiffness matrices of the laminate, as
follows,

|A] = ] |

]Jz
h
t
-h
b
|B] = ] z|

]Jz
h
t
-h
b
|B] = ] z
2
|

]Jz
h
t
-h
b
(2. 80)

where [A], [B] and [D] are the stiffness matrices of the laminate and |

] the stiffness matrix


of the ply located at the considered altitude. As |

] is constant in a given lamina, these


integrals can be written as a summation as follows:

A
Ij
= (

]
)
k
(z
k
-z
k-1
)
N
k=1
i, ] = 1, 2, 6 (2. 81)

B
Ij
=
1
2
(

]
)
k
(z
k
2
-z
k-1
2
)
N
k=1
i, ] = 1, 2, 6 (2. 82)

B
Ij
=
1
3
(

i]
)
k
(z
k
3
-z
k-1
3
)
N
k=1
i, ] = 1, 2, 6 (2. 83)

Where N is the total number of plies in the laminate, z
k
and z
k-1
are z-coordinates of the two
surfaces of the ply number k and (

]
)
k
are the elements of the stiffness matrix in the
Cartesian coordinate system (x, y, z) of the ply number k. Then, the equations 2.78 and 2.79
can be merged in the following expression:

`
1
1
1
1
1
1
N
x
N

N
x
H
x
H

H
x1
1
1
1
1
1
1
= _
|A] |B]
|B] |]
_
`
1
1
1
1
1
1
e
x
o
e
y
o
y
xy
o

x1
1
1
1
1
1
1
(2. 84)

That represents the generic behavior law for a laminate. Being A
ij
the in-plane stiffnesses that
relate the in-plane forces N
x
, N
y
, and N
xy
to the in-plane deformations e
x
o
, e

o
and y
x
o
. The D
ij

2 Theoretical Background 30


are the bending stiffnesses that relate the moments M
x
, M
y
, and M
xy
to the curvatures
x
,


and
x
. Finally the B
ij
, are the out-of-plane coupling stiffnesses that relate the in-plane forces
to the curvatures and the moments to the in-plane deformations.
2.2 Composite Failure Theory
The failure in a mechanically fastened joint occurs as a combined and inter-dependent failure
of each individual component that is not only the joined surface (plate, laminate), but also
the joining element (bolt or rivet). This makes prediction very difficult due to all the
parameters that have to be taken into account. To estimate the failure of joints in numerical
models separate models are used for the fastener and the composite. As the fastener is usually
made of metal the failure prediction is relatively easy and relies on the ultimate strength or
ultimate elongation of the fastener. But for laminates the failure prediction is very complex
due to the characteristics of the material. Failure can occur at different levels, in the fibers, in
the matrix or between the layers, and it occurs progressively.
The failure initiation prediction begins by predicting the failure of the elementary lamina
composing the laminate, based on failure criteria. Then the parameters are adjusted according
to the results of the ply analysis and validated against physical material tests.
2.2.1 Failure of a Lamina
2.2.1.1 Failure Modes of a Laminate
As shown in Figure 2.18, there are four main failure modes of a laminate: microbuckling,
fiber breakage, matrix cracking and delamination. The first three are microscopic failures, and
the last is a macroscopic one.

Fig. 2.17: Typical failure modes of composites.
2.2.1.1.1 Microbuckling
Microbuckling or local fiber buckling reduces the compressive stiffness and strength of the
laminate. Microbuckling does not necessarily lead to immediate failure because the
surrounding matrix supports the fibers. The properties of the fibers and the matrix greatly
2 Theoretical Background 31


affect the onset and magnitude of fiber buckling and the resulting losses in the compressive
properties of the laminate.
2.2.1.1.2 Fiber Breakage
Fiber breakage refers to the failure of the fiber under a longitudinal tensile load. In order to
confront tensile loading composite materials have fibers. Advanced composites are commonly
composed of highly elastic fibers with high tensile strength. Then if the fibers fail, the load is
redistributed by the matrix to the surrounding fibers. The matrix acts as a bridge, transmitting
the load across the gap created by the breakage. Even with a certain amount of broken fibers,
the laminate should be able to confront the loading.

2.2.1.1.3 Matrix Cracking
Matrix cracking may occur under any loading. This can be a serious issue if several fibers are
already broken. The laminate could as well be weakened due to an increase in the moisture
absorption, reducing the stiffness of the laminate and the crack could be propagated into the
interface between adjacent layers, initiating delamination.

2.2.1.1.4 Delamination
The delamination consists in the separation of two adjacent layers. It can occur because of
manufacturing defects, edge defect, and impact by foreign objects. This failure is not easy to
predict and detect by a visual inspection because it often occurs within the laminate. The
result is a reduction in the bending stiffness and strength, as well as the load carrying capacity
of the laminate under compression.
To predict these phenomena theoretically, failure criteria are used.
2.2.1.2 Failure Criteria
Failure criteria are analytical functions aiming at predicting the failure onset of a lamina. They
are applied successfully to every ply constituting the laminate to determine the initiation of
the failure. Failure criteria have usually the following form:

F(o
1
, o
2
, o
3
, z
23
, z
13
, z
12
, F
1
, F
2
, ) _
< 1 nu a||ure
= 1 a||ure ||m|t
> 1 a||ed
(2. 85)

In the expression, the stresses are the stresses in the ply and F
1
, F
2
are the strength
parameter of the ply for different kind of loading.
2 Theoretical Background 32


2.2.2 Maximum Value Failure Theories
Maximum value failure theories are divided in two categories, one using stress
considerations and other using strain considerations. These criteria belong to a structural
type and are based on the assumption that there can be three possible modes of failure
caused by stresses
1
,
2
,
12
or strains
1
,
2
,
12
when they reach the corresponding
ultimate values. This seems like a direct approach to evaluate the failure of a material.

2.2.2.1 Maximum Stress Theory
This theory predicts the failure of the ply when any component of the stress tensor exceeds
the strength of the material for this solicitation, that means in a compressive, tensile or shear
state. The direction 1 is thus along the fiber direction, the direction 2 in the lamina plate,
normal to the first one and the direction 3 normal to the lamina plane. The failure thus occurs
when the following relation is verified

mox [
c
1
X
t
,
c
2

t
,
c
3
z
t
,
c
1
X
c
,
c
2

c
,
c
3
z
c
,
:
23
S
23
,
:
13
S
13
,
:
12
S
12
1 (2. 86)

where X
t
, Y
t
, and Z
t
are the tensile strengths and X
c
, Y
c
, and Z
c
are the compressive strengths
of the ply in the directions 1, 2 and 3 respectively and S
ij
are the shear failure strengths in the
plane (ij).
2.2.2.2 Maximum Strain Theory
Using the same scheme with E
1t
, E
2t
, and E
3t
are the tensile strengths and E
1c
, E
2c
, and E
3c
are
the compressive strengths of the ply in the directions 1, 2 and 3 respectively and E
ij
are the
shear failure strengths in the plane (ij), the following equation can be drawn:

mox [
s
1
L
1t
,
s
2
L
2t
,
s
3
L
3t
,
s
1
L
1c
,
s
2
L
2c
,
s
3
L
3c
,
y
23
L
23
,
y
13
L
13
,
y
12
L
12
1 (2. 87)

2.2.3 Interactive Failure Theories
Interactive criteria are criteria predicting the failure load by using polynomial equations
involving stress components. As all the loads are taken into account at the same time, loads in
any direction can influence the failure in every other direction. The two common interactive
failure theories are the Tsai-Wu and Tsai-Hill.
2.2.3.1 Tsai-Wu Theory
The Tsai-Wu theory is a tensor polynomial theory predicting the failure onset of a ply using
the following second-order-tensor polynomial equation:
2 Theoretical Background 33



F

+ F
]
o

o
]
6
=1
6
=1
6
=1
1 i, ] = 1,2, ,6 (2. 88)

The coefficient F
i
and F
ij
are the strength parameters, functions of the tensile, compressive or
shear strengths of the material.
For example, in case of plane stress state, the equation 2.74 can be simplified and the failure
is initiated when:

F
1
o
1
+F
2
o
2
+F
11
o
1
2
+2F
12
o
1
o
2
+F
22
o
2
2
+F
66

12
2
1 (2. 89)

In which the strengths are given by:

F
11
=
1
X
t
|X
c
|
F
12
=
1
X
t
|X
c
|
+
1

t
|
c
|
F
22
=
1

t
|
c
|
F
1
=
1
X
t
-
1
|X
c
|
F
2
=
1

t
-
1
|
c
|
F
66
=
1
S
12
2
(2. 90)

Where like previously, X
t
, Y
t
and Z
t
are respectively the tensile strengths of the ply in the
directions 1 and 2, X
c
and Y
c
are the compressive strengths in the directions 1 and 2, and S
12

is the shear failure strength in the plane (12). One can notice that the parameter F
12
is not
considered. This parameter is usually determined experimentally by a biaxial test although
some formulas exist to try to predict it.
Tsai and Wu originally insisted on the need of maintaining closed, ellipsoidal failure surfaces
for all stress states. They proposed stability conditions on that purpose like the inequality:

F

F
]]
-F
]
2
u (2. 91)

The summation is not implied by repeated indices. In the case of plane-stress conditions in the
plane (12), thus:

F
11
F
22
-F
12
2
u (2. 92)

The equation 2.78 is the equation of an ellipse. The failure surface is illustrated in Figure 2.19
for F
12
= u. One can notice in equation D.8 the influence of F
12
which entails the rotation of
the failure surface.
2 Theoretical Background 34


2.2.3.2 Tsai-Hill Theory
Tsai-Hill theory is the application of Hills anisotropic plasticity theory to the failure of
heterogeneous, anisotropic materials. In plane-stress state in the plane (12) the failure occurs
when:

F
11
o
1
2
+2F
12
o
1
o
2
+F
22
o
2
2
+F
66

12
2
1 (2. 93)

Where the strength parameters are:

F
11
=
1
X
2
F
12
= -[
1
X
2
+
1

2
-
1
z
2
F
22
=
1

2
F
66
=
1
S
12
2
(2. 94)

With X, Y and Z equal to X
t
, Y
t
and Z
t
or X
c
and Y
c
depending on the sign of
1
and
2
and
S
12
of the shear strength in the plane (12).
Note that in the case of plane stress in 12, for a transversely isotropic laminate ( = Z), so
F
12
=
-1
X
2


can be considered. This is the quadratic criterion.
All these interactive theories brought an enhancement compared with the maximum value
theories by considering the interaction between the stresses. Composite materials are
heterogeneous and furthermore anisotropic materials so failure does not occur the same way
in the same direction, something that is not accounted for these theories.

Fig. 2.18: Failure surfaces for some failure criteria. [9]
2 Theoretical Background 35


2.2.4 Mechanical Failure Theories
The mechanical failure criteria are nowadays widely used because they enable to predict the
failure of the lamina by identifying the failure mode as well as which loading state is
responsible for this failure.
2.2.4.1 Hashin-Rotem Criterion
Hashin-Rotem failure criterion is a 2D criterion, one of the firsts to predict the failure mode in
composite materials. This criterion is applied assuming plane stress conditions in the plane
(12). The equations are as follows:
Tensile fiber failure mode:

c
1
X
t
= 1 (2. 95)

Compressible fiber failure mode:

c
1
X
c
= 1 (2. 96)

Tensile matrix failure mode:

[
c
2

2
+[
:
12
S
12

2
= 1 (2. 97)

Compressible matrix failure mode:

[
c
2

2
+[
:
12
S
12

2
= 1 (2. 98)

Where X
t
and Y
t
are respectively the tensile strengths in the fiber direction and in the
transversal direction, X
c
and Y
c
are the compressive strengths in the fiber direction and in the
transversal direction and S
12
is the shear failure strength in the plane (12). It can be seen that
the Hashin-Rotem criterion is a further development to the Tsai-Hill criterion as an enhanced
failure criteria.
2 Theoretical Background 36


2.2.4.2 Hashin Criterion
The Hashin criterion comes from the Hashin-Rotem criterion. Indeed, this is an extension of
the previous criterion to three dimensional problems. Now the equations are as follows:
Tensile fiber failure mode:

c
1
X
t
= 1 or [
c
1
X
t

2
+
1
S
12
2
(
12
2
+
13
2
) = 1 (2. 99)

Compressive fiber failure mode:

c
1
X
c
= 1 (2. 100)

Tensile matrix failure mode (o
2
+o
3
u):


1

t
2
(o
2
2
+o
3
2
) +
1
S
23
2
(
23
2
-o
2
o
3
) +
1
S
12
2
(
12
2
+
13
2
) = 1 (2. 101)

Compressive matrix failure mode (o
2
+o
3
< u):

1
|
c
|
_[

c
2S
23
2

2
-1_ (o
2
2
+o
3
2
) +
1
4S
23
2
(o
2
2
+o
3
2
) +
1
S
23
2
(
23
2
-o
2
o
3
) +
1
S
12
2
(
12
2
+
13
2
) = 1
(2. 102)

where it is assumed that S
12
= S
13
and where S
23
is the shear strength in the plane 23. For the
2D case, the failure criteria can be obtained easily by getting rid of every values accounting
for the direction 3.
2.2.5 Delamination Initiation Prediction
Most of the delamination onset criteria used are stress based criteria, some hereafter
presented. Relying on the same principle as the other criteria, Pearce [10] lists in his papers
some stress based criteria; it is based on the comparison to allowable stress values.
The first natural criterion is:

2 Theoretical Background 37


c
3
z
t
1 (2. 103)

where Z
t
is the through-the-thickness tensile strength.
This criterion only considers the through-the-thickness load which can, in a first approach,
give good results.
Ye [11] proposed another criterion, which includes an interaction between normal and shear
through-the-thickness stress. It is expressed with the two following equations.
Tensile delamination mode (o
3
u):

[
c
3
z
t

2
+[
:
13
S
13

2
+[
:
23
S
23

2
1 (2. 104)

Compressive delamination mode (o
3
< u):

[
:
13
S
13

2
+[
:
23
S
23

2
1 (2. 105)

where S
13
is the in-plane shear strength, S
23
is the transverse shear strength and Z
t
is the
through-the-thickness tensile strength.
This criterion differentiates two modes, a tensile and a shear mode.
Tensile delamination mode:

c
3
z
t
1 (2. 106)

Shear delamination mode:

:
13
2
+:
23
2
S
23
2
1 (2. 107)

2.2.6 Damage Progression Models
The failure of some plies within a composite laminate does not necessarily lead to the collapse
of the structure. The laminate can sometimes withstand the loading. The failure onset
prediction of a ply is, as a result, not sufficient to predict the ultimate failure of the laminate.
2 Theoretical Background 38


Damage progression models are used in order to take into account the progressive material
properties degradations. These models alter the stiffness or compliance coefficients ruling the
elastic constituting relations in a laminate.
There are two main types of degradation models. On one hand, the models are based on ply-
discounting material degradation approach. On the other hand, the models are based on a
continuum-damage-mechanics approach using internal state variables.
2.2.6.1 Ply-Discounting Approach
The ply-discounting approach is a common approach for the failure analysis. It is taken the
consideration that the laminate behaves as a homogeneous material. The laminate plate theory
is used to calculate the stresses and strains in each ply. Then it is analyzed with a lamina
failure criterion in order to determine which ply fails first and possibly the mode of failure. A
stiffness reduction model is then applied to decrease the stiffness of the laminate due to this
ply failure.
The stiffness reduction model relies on the application of decreasing coefficient to the elastic
material stiffness coefficients affected by the ply failure. There are several stiffness-reduction
models like the parallel springs, distinct for the fibers and the matrix or the incremental
stiffness reduction mode [17].
Due to its easy implementation, this approach is widely used. Using this method, the
predictions can however either under- or over- predict the stiffness decrease because of matrix
damage depending on the stacking sequence.
2.2.6.2 Continuum-Damage-Mechanics Approach
Continuum-damage-mechanics (CDM) models generally describe the internal damage in the
material by defining one or more internal state variables [17]. The CMD models use similar
damage law relationships as the elastic law relationships to describe the behavior of the
laminate, except for the coefficients of the stiffness or compliance matrices, which depend on
a variable describing the damage state of the material (noted usually as ). These models can
also be used to predict the delamination propagation and can be implemented through the
finite element analysis.
2.3 Composite Bolted Joints in Aircraft Structures
2.3.1 Mechanically Fastened Joints
2.3.1.1 Generalities
Fastening consists of connecting two distinct parts together in order to create an assembly or a
structure capable of sustaining a certain loading. The joint created has to be able to transfer
the loads between the different parts. In a structure, a joint is the weaker area and most
failures initiate around the joints. Because of this, joints are sometimes eliminated by
2 Theoretical Background 39


integrating the structure. As used by K. Mazumdar [12], joints have the following
disadvantages:
A joint is a source of stress concentration. It creates discontinuity in the load transfer.
The creation of a joint is a labor-intensive process; a special procedure is followed to
make the joint.
Joints add manufacturing time and cost to the structure.
There are several techniques for joining components of fiber-reinforced composites,
mechanical fastening and adhesive bonding and fusion or melting.
Mechanical joining is most widely used in joining metal components. Examples of
mechanical joints are bolting, riveting, screw and pin joints. Similar to the mechanical joints
of metal components, composite components are also joined using metallic bolts, pins, and
screws; except in a few cases where RFI shielding and electrical insulation are required,
composite fasteners are used.
The main advantages of mechanical fastening are:
No abnormal inspection problems
Disassembly possible without component damage
No surface preparation of components required
The main disadvantages of mechanical fastening are:
Holes cause unavoidable stress concentration
Can incur a large weight penalty
For most mechanical joints, an overlap is required in two mating members and a hole is
created at the overlap so that bolts or rivets can be inserted. When screws are used for
fastening purposes, mostly metal inserts are used in the composites, the reason being that the
threads created in the composites are not strong in shear and therefore metal inserts are used.
In bolted joints, nuts, bolts and washers are used to create the joint. In riveting, metal rivets
are used. Bolted joints can be single lap joints, double lap joints or butt joints.
2.3.1.2 Failure Modes of Mechanically Fastened Joints
Experimentally determined behavior of both small and full-size joints is additionally required
to complement the theoretical work. Knowledge of failure modes at macroscopic and
microscopic levels is essential, if theoretical models are to be refined. One of the problems
with composites is the very large number of material that can be evolved by changing fiber,
matrix, lay-up, stacking sequence, etc. Taking into account the huge cost of testing all
possible combinations, as far as a general predictive method, which does not exist until date,
the detailed three-dimensional analyses show encouraging correlation between stress
distribution and experimentally observed failure modes.
2 Theoretical Background 40



Fig. 2.19: Modes of failure for bolted joints in advanced composites.
2.3.1.2.1 Tension Failure
It is also known as Net section failure, the tensile load required to fail a laminate through a
section at which holes occur (net section) is less than at a section at which there are no holes
(gross section), as is shown in Figure 2.16. The failure stresses at these sections can be
represented by

o
N
=
P
(w-N)t
(2. 108)

And

o
u
=
P
wt
(2. 109)

where P is the critical load, W the width at the net section of the joint, N the number of holes
of D diameter at that section, and t the joint thickness. This failure is one of the most
dangerous modes because it is sudden and occurs without any early warning.
2.3.1.2.2 Shear Failure
The shear strength is the interlaminar shear strength obtained using the well-known short
beam shear test technique [13]. The shear strength, as for isotropic materials, is given as

2 Theoretical Background 41


=
P
2Lt
(2. 110)

where E is the distance (parallel to the load) between the center of the hole and the free edge,
usually known as end distance. The joint shear strength of unidirectional composites, when
loaded in the direction of the fibers, is low in comparison with the in-plane shear strength.
This mode is common for highly orthotropic laminates.
2.3.1.2.3 Bearing Failure
The bearing failure is a more progressive failure mode that does not entail a maximum
indicating the beginning of the bearing failure, but which is then still able to carry the load
until the complete failure of the joint in another failure mode like pull-through or shear-out.
For practical purposes, the bearing strength of a composite material is usually expressed as an
average stress acting uniformly over the cross-sectional area of the hole, so that

o
b
=
P
Nt
(2. 111)

The point of failure will depend upon the fiber orientations and the proportion of each
orientation contained in the laminate. It has also been shown that laminates containing some
45 and/or 90 plies perform well under bolt bearing conditions although this is contrary to
what might be expected from knowledge of the compressive strength behavior of 45 and
90 laminates, due to the lateral constraint provided by clamping action of the bolt.
2.3.1.2.4 Cleavage Failure
Cleavage failure occurs only in u _0 lay-ups which contain low proportions of non-axial
fibers. Failure initiates in a single shear mode, this is then followed by the failure of the net
section on one side of the laminate. For most applications, lay-ups containing fiber orientation
that are susceptible to this type of failure would not be used. [14]
2.3.1.2.5 Pull-out Failure
A composite bolted joint in single lap-shear configuration undergoes a very complex failure
and damage mechanism. In most cases, the first failure load is reached with the onset of inter-
laminar cracks, leading to bearing failure and elongation of the hole, and finally a pull-out
failure of the bolt.
2 Theoretical Background 42


2.3.2 Design Considerations: Material Parameters and
Joint Geometry
2.3.2.1 Fiber Orientation
The influence of the fiber orientation and the proportion of fibers at the various orientations
can best be demonstrated by looking at single bolted joint in a particular family of laminate
orientations. As has been shown by Collings [15], the best overall joint performance for
reinforced composite structures is achieved by using a combination of 0 and 45 plies,
which strongly influences both the material property and the joint behavior.
2.3.2.2 Stacking Sequence
The stacking sequence has a significant impact on the joint behavior. Most of the time, the
preferred configuration for composite plies is the quasi-isotropic configuration composed of
0, 45, 90 fibers [16]. This kind of composite material is indeed more likely to be able to
sustain the complex loading in a joint. The non-zero degree fibers carry the load around the
hole to prevent shear-type failures, whereas the 0 fibers carry out the primary bearing load
and tension. The stacking sequence has to be chosen according to the desired failure mode.
Park [17] studied the influence of the stacking sequence for different quasi-isotropic
composite laminates |9u
3
_4S
3
u
3
]
s
, |u
3
_4S
3
9u
3
]
s
, |9u
3
u
3
_4S
3
]
s
with and
without clamping, pointing out that in both cases, the trend of variations for the ultimate and
delamination bearing strength of bolted joints for the same stacking sequence were the same,
but clamping allowed to reach significantly higher strength of the laminate.
2.3.2.3 Joint Dimensions
The geometry has also a critical influence on the failure of the joint. The effects of specimen
geometry on joint life were eva1uated for the 50/40/10 and 19/76/5 lay-ups using specimens
with different edge distances and widths. Fatigue strength was not changed for the 50/40/10
lay-ups. For the matrix-dominant 19/76/5 lay-up, fatigue strength was found to increase when
the w/d ratio was increased from 3 to 4. Further, a change of failure mode from bearing to net
section occurred when w/d ratio was reduced from 4 to 3.
2.3.3 Design Considerations: Fastener Parameters
2.3.3.1 Clamping Torque
The procedure for installation of fasteners in composite structures can produce unacceptable
damage. Before the installation of fasteners, holes must be cleanly drilled through the
composite structure. This is not an easy task because composites are prone to fraying within
the hole and to splintering on the exit side of the hole as a result of the drilling
operation; consequently, special drill bits and procedure different from those used in
metal drilling, are required in order to avoid introducing any local weakening of the
composite structure. While the fastener is being installed into the hole, its rotation is
not recommended because this would lead to breaking and lifting of fibers from the
2 Theoretical Background 43


surface. Also, axial misalignment of fastener in the hole could lead to damage. For example,
misalignment of a blind fastener during installation could cause the blind head to dig into the
surface of the hidden side, crushing or delaminating the composite skin. Several
manufacturers have introduced fastener designs which have significantly facilitated their
installation in composite structures.
2.3.3.2 Friction
The effect of friction is commonly ignored in the analytical work published in the literature.
Friction between plate surfaces can, however, significantly affect joint bolt load distribution.
Experimental work by Wittmeyer and Smode [18] indicates that the clamp-up force resulting
from bolt tightening relieves the joint load transmitted by fastener shear. Significant fretting
was observed in the region of failure. This observation led to the belief that failure was the
result of the propagation of cracks which were initiated by a fretting mechanism. In
composites, this contact problem in the faying plate surfaces is further complicated by the fact
that the behavior of friction and wear is a function of varying fiber orientations with respect to
the sliding direction.
2.3.3.3 Bolt/Hole Clearance
The drilling is a critical task in composite materials because, drilling defects can produce a
significant effect on the strength of the joint, inducing additional damage to the material,
actually, a poor manufactured hole can produce the 60% of all part rejections.

Fig. 2.20: Schematics of delaminations caused by drilling (a) upon entry, and (b) upon exit. [19]
Drilling entails the weakening of the material by triggering a discontinuity in it but moreover,
delamination in the outer layers can occur during manufacturing process, as it is shown in
Figure 2.17.
2.4 Finite Element Modeling
Finite element methods are taking an increasing role in the analysis and design of complex
structures. With the development of many finite element codes and the increase of
computational power, it becomes indeed easier to create a good model of an ARCAN test and
to simulate it. It also appears as a way of saving time by reducing or even sometimes, for
really accurate model, replacing to a great extent experiments. It enables, as a result, to save
time and cost of developments. This was precisely the purpose of the already mentioned
European program BOJCAS [20] for bolted joints. In the frame of this project, researchers
2 Theoretical Background 44


tried to get a better understanding of the behavior and the mechanisms occurring in bolted
joints by using of course experiments and also using finite elements analysis. From their
work, they have been able to develop a complete tool for the design of bolted joint finite
elements models called Abaqus [21].
In this section, after having given the main equations of the finite elements method used
hereafter, we will give a short overview of some works made in bolted joints modeling and
we will finally describe the main damage modeling method in finite element analysis.
2.4.1 Finite Element Theory
2.4.1.1 Hypothesis of the Finite Element Analysis
Mechanically fastened joints made of reinforced composite material have many specificities
that make its analysis complicated. Three-dimensional and singular phenomena have to be
modeled or determined to obtain a realistic model of the joint. Among these, the bolt
clamping force and thus the bolt pretention, the bearing stresses, the stress repartition at the
bolt hole interface, the delamination or also the secondary bending need to be mentioned, all
these mechanisms interacting.
Other parameters like contact and friction oblige to take into account non-linearity effects. As
mentioned section 2.2, the bolt-hole contact, the washer-composite contact or the contact
between the composite plates can greatly influence the stress repartition.
The properties of these contact zones, like the friction definition or note, have also to be
carefully addressed. It will be specified as it is part of a crash study. Short-period phenomena
have thus to be modeled. Given all these considerations, explicit finite elements formulation
was used for this study.
2.4.1.2 Explicit Finite Element Analysis
According to the theory of the dynamic mechanics, the motion equation of a solid undergoing
damping in a Lagragian coordinate system is given by [22]:

|H]{q ]
n
+|C]{q ]
n
+|K]{q]
n
= {R
cxt
]
n
(2. 112)

Where {q]
n
is the array of the lagrarian generalized coordinates of the solid at the n-
increment, {q ]
n
and {q ]
n
are respectively their first and second order time derivative, [M] is
the mass matrix of the solid, [C] the damping matrix, [K] the stiffness matrix and {R
cxt
]
n
the
array of the exterior forces applied on the solid at the n increment.
There are two main techniques to solve this dynamic equation including non-linearity.
An implicit method so-called standard or an explicit method can be used. The first one
consists in the determination of {q]
n
at t
n
using the current state time derivative {q ]
n

2 Theoretical Background 45


and {q ]
n
, using thus a direct integration scheme. For this method, the commonly used
numerical integration techniques are the Newton or Newton-modified techniques [22].
The other method is the explicit finite element analysis method which is based on the explicit
central-difference integration rule:

u
[+
1
2

N
= u
[-
1
2

N
+
t
(i+1)
+t
(i)
2
u
()
N
(2. 113)

u
(+1)
N
= u
()
N
+t
(+1)
u
(+
1
2
)
N
(2. 114)

Where u
N
designates a degree of freedom, a translation or a rotation, and the subscript i
refer to the increment number in an explicit dynamic step. This operator is said to be explicit
in the sense that the kinematic state is advanced using the values of the previous
increment u
[-
1
2

N
, u
()
N
.
The main advantage in the use of this integration method over the implicit finite element
analysis method is the use of a diagonal or lumped mass matrix [M]. There are several
procedures to get a diagonal matrix, one common being the row/sum technique, in which the
diagonal elements of the mass matrix [23] are obtained by the summation of the elements of
the each row of the consistent mass matrix law. The inversion of the mass matrix is thus easy
and quick for the explicit method, the updating of the displacements do not require the
solution of complex equations and no tangent matrix has to be formed.
The main disadvantage of the explicit finite element method is the stability of the method.
The implicit method is intrinsically stable. The increment stable time can be relatively high
compared to the increment stable time of the explicit method. The stability of this last one is
ruled by the time step of each increment that has to be below a critical time increment value
so that:
t
ct
=
L
min
c
d
(2. 115)
Where I
mn
is the smallest element dimension in the mesh and c
d
is the dilatational wave
speed defined in function of the Lame coefficients , and the density of the material, as
follows [22]:
c
d
= _
x+2q
p
(2. 116)
Given this limitation, even if the computational cost of each increment is relatively low, it is
required to get to the end of the analysis.
2 Theoretical Background 46


For large problems, the explicit method is however faster than the implicit method because for
the first one, analysis cost rises linearly with the problem size, whereas for implicit integration
the cost of solving the nonlinear equation rises faster than linearly with problem size.

2.4.2 Bolted Joints Finite Element Modeling
The finite element modeling is a technique of analysis of structures more frequently used,
given the development of performing finite element tools and of the increase of the
computational power. It is also as a consequence used to predict the behavior of mechanically
fastened joints, enabling to cut experimentation time cost. Two types of mechanically
fastened joints can be distinguished, the pinned joints and the bolted joints.

2.4.2.1 Pin joints Modeling
Pin joints were subjected to many studies [24]. The first finite element models of pinned
joints developed were 2D models. They are interesting for the few computational capacities
they required. They provide reasonable prediction of the failure load as well as of the failure
initiation for large geometric ratios E/D and W/D [19]. The pin load is usually simulated
either by applying a radial boundary condition to the nodes around the whole plate, or a co
sinusoidal radial stress on the loaded side of the hole. A more accurate method is however to
model the pin load by modeling the pin as a rigid surface and defining a contact surface
between the pin circumference and the nodes around the hole [25].
The 2D models cannot however account for the out-of-plane phenomena like delamination, or
the effect of the stacking sequence. As a consequence, even if the 3D model requires more
work to develop, they can theoretically provide more accurate models capable of predicting
the failed load initiation of the joint and its failure mode, using more advance 3D failure
criteria. Some models can achieve good correlations with the experimentations [19].
The study of pinned joints is also sometimes prior to the study of the gains of clamping in
bolted joints.

2.4.2.2 Bolted Joints Modeling
The study of bolted joints requires taking into consideration more complicated phenomena
compared with the secondary bending or 3D stress repartition within the material. They have
been the subject of many studies and especially in the last years. In the different models
developed, the authors try to account for the failure mechanism, for some precise mechanism
like the bolt-hole clearance or the bolt tilting effect [26], the effect of secondary bending [27],
the effect of the clamping force or the effect of the geometric parameters of the joint [28], the
behavior of multi-bolted joints [29] Most of the time these models are able to describe
relatively well the behavior of the joints under some assumptions and regarding the particular
effect they are focusing on. They can give a realistic prediction of the failure of the bolt. The
2 Theoretical Background 47


choice of the failure criteria and of the damage progression models appears to be as well
critical parameters. They have to be carefully chosen.
A relatively high degree of accuracy can be reached with the 3D model of bolted joints. This
is however at the expense of the computational efficiency of the model. The modeling of the
interlaminar damage for example can oblige a high refinement of the model, entailing the
increase of the calculation time or power needs.
In the following section the main techniques used to model the complex phenomena of
interlaminar damage within a composite material will be presented.

2.4.3 Interlaminar Damage Modeling
Delamination is a common form of failure in laminates owing to the lack of reinforcements in
the thickness direction. Being able to model and predict this phenomenon, consisting in the
propagation in cracks within the matrix in the fibre-plane, is thus important to have a reliable
model. The delamination onset is predicted using delamination onset criteria like the one
mentioned in section 2.3.5. Several methods exist for the simulation of the delamination
growth like the J-integral method, the Virtual Crack Closure Technique, the Cohesive Zone
Method [30]. Most rely on the fracture mechanics method. The two mainly used methods,
the VCCT and the CZM methods, will be presented here.

2.4.3.1 Fracture Mechanics Theory
2.4.3.1.1 Griffith Energy Criterion
The fracture mechanics theory enables to characterize the crack propagation. It can as a result
be used to monitor the delamination in a laminate as the mechanism is similar to crack
propagation. The Griffith energy criterion can thus be applied. If the energy release rate is
equal or larger than the energy required to propagate the crack surface, called the critical
energy released rate [31]. The energy release G
c
is usually defined for a double cantilever
beam specimen loaded, in with crack propagates of a length a the energy rate is defined as
[32]:

0
C
=
1
b
(
d0
cxt
du
-
d0
int
du
) (2. 117)

Where u
cxt
and u
nt
are respectively the external energy due to the external forces P applied
on the beam and the energy stored in the two arms of the DCB, b is the width of the material
releases per change in unit crack growth per unit depth under a solicitation mode.
The critical energy release rate is generally separated in three components. Each component is
relates with a different loading case, in a different direction and therefore to a different mode
2 Theoretical Background 48


of propagation of the crack. The critical energy release rate for every mode is determined
experimentally using mainly DCB test specimen.

Fig. 2.21: Double Cantilever Beam undergoing crack growth. [32]


Fig. 2.22: Fracture mechanics crack propagation modes.
2.4.3.1.2 Mixed-Mode Delamination
A structure does not usually fail purely in a single mode but rather in a combination of these
modes called mixed mode, to account for the interaction between the different modes, several
criteria have been developed like the BK law, the power law, and the Reeder law models. The
generic form of these criteria can be defined as follows, using the following ratio between G
ep

2 Theoretical Background 49


the equivalent strain energy release rate calculated at the crack tip and G
eqc
the critical
equivalent strain energy release rate

u
cq
u
cqc
1 (2. 118)

To take into account the interaction between the modes several delamination growth criteria
have been developed. The mentioned criteria are widely used. It enables to combine the
influence of the three modes at the delamination tip. It is expressed as follows:

[
u
I
u
IIc

u
+[
u
II
u
IIc

[
+[
u
III
u
IIIc

y
1 (2. 119)

Where G
I
, G
II
and G
III
are respectively the energy release rates in mode I, II, and III and G
Ic
,
G
iiC,
and G
IIIc
are the corresponding critical energy rates. The parameters , , y are most of
the time chosen all equal to 1, which is the linear failure criteria, or all equal to 2, which
corresponds to the quadratic failure criteria.
In practice, it is difficult to distinguish the modes II and III. They are generally combined into
a single resultant shear deformation. T he power law criteria becomes as a result

[
u
I
u
Ic

u
+[
u
shccr
u
Sc

[
1 (2. 120)

Where G
shear
is the energy release rate in shear mode and G
Sc
is the corresponding critical
energy release rate and here as well commonly ==1 or 2. Fracture mechanics method is
implemented in some finite elements codes. This is for example used in the virtual crack
closure technique in the finite element software Abaqus.

2.4.3.2 Virtual Crack Closure Technique
The VCCT is a commonly applied method for determining the components of the strains
energy release rate defined in the previous section along a crack front. It is based on two main
assumptions. The first one, called the Irwins assumption, assumes that its original length, and
the second one, supposes that growth does not significantly alter the state at the crack tip [31]
before the propagation of the crack and after, as shown in Figure 2.22 (left).
The components of the energy release rate can be determined using different expressions. To
give an example, the Figure 2.22 of a DCB under pure Mode I, will be considered. By
2 Theoretical Background 50


assuming that the crack closure is ruled by linear elastic behavior, the energy needed to close
the crack and, therefore the energy to open the crack, is given by:

0
1
= -
1
2
P
]

i
ub
(2. 121)

Where F
]
, :

, o and b are respectively the node reaction force and j to the crack closure, the
vertical displacement between nodes at I, the crack length extension and the width of the
bean. Applied to the case in Figure 2.22, the delamination will propagate at the nodes 2 and 5
when:

0
Ic
0
1
= -
1
2
P
]2,S

1,6
db
(2. 122)

With 0
Ic
the mode I fracture toughness parameter is experimentally determined. Similar
considerations can be made for the other modes to determine the energy release rates. One
requirement of the VCCT method is the initial definition of the delamination, especially the
location of the delamination front, what, in some cases, may be difficult to determine. This is
however not the case for the cohesive zone models.

2.4.3.3 Cohesive Zone Models
2.4.3.3.1 Principle
The cohesive zone models offer the advantage of predicting both initiation and growth of
delamination, the first one, using a strength criterion and the second one, the separation
prediction, is governed by fracture mechanics parameters [31]. They assume that it is possible
to introduce interface elements, where the delamination can occur and propagate (thus
typically between the laminate plies), enabling the modeling of the propagation of the
delamination by the introduction of a discontinuity in the displacement field. These elements
called, cohesive or process zones, cohesive elements or interfacial de-cohesion elements are
based in the Dugdale-Barenblatt type cohesive zone and where first used for prediction of
concrete cracking [33].
2 Theoretical Background 51



Fig. 2.23: VCCT approach for pure Mode I: Crack Extension (left), Crack Growth Criterion (right). [22]
An example of an interfacial decohesive element inserted between the two arms of a DCB is
showed in Figure 2.23 they are particularly interesting when the interfacial strength is
relatively weak compared with the adjoining material, as it is the case for composite laminate.

Fig. 2.24: Cohesive Zone in a double cantilever beam. [34]
These materials are conferred a specific behavior. The de-cohesion response is specified in a
function of the interfacial tractions and the relative tangential and normal displacements
across the interface. The work resulting of normal separation and tangential separation can be
related to the critical values of energy releases rates. Figure 2.25 shows an example of a
traction separation law for a double cantilever bean, loaded under Mode I with every
noticeable area of the process zone indicated on the constructive curve. The material is still
intact regarding the delamination before crack tip at the point 2, where the stress state is the
highest. Further from the crack tip a softening occurs. The cohesive zone can still transfer
load after delamination onset at the point 3, until the critical value of the energy release rate is
attained at the point 4. After the point 4 the cohesive zone has completely failed. Notice that
this de-cohesion element uses high initial stiffness before delamination onset to prevent
additional deformations (see point 1), what is generally the case.
2 Theoretical Background 52



Fig. 2.25: Cohesive Zone ahead of delamination tip. [34]
Considering the constitutive law in Figure 2.25, the energy release rate can be expressed as
follows:

u = ] ()u
8
t
0
(2. 123)

Where
t
is the relative displacement at the crack tip.
The critical energy release rate G
C
resulting in the crack propagation is reached when the
displacement at the crack tip is maximum, when
t=

max
. It can be written as follows:

u
C
= ] ()u
8
max
0
(2. 124)

The values of G and G
C
are determined by the constitutive equation in the formulation of the
constitutive elements.
2.4.3.3.2 Constitutive Equations
The constitutive response of the de-cohesion elements is modeled based on a traction-
separation description of the interface. As is shown in Figure 2.26, many equations can be
used to describe the damage evolution:
Linear softening.
Perfectly plastic.
2 Theoretical Background 53


Progressive softening.
Regressive softening.
Needleman.
The critical energy release rate is determined as the area under the constitutive curves.

Fig. 2.26: Constitutive strain softening equations. [34]
An initially high stiffness is used to hold the top and the bottom faces of the de-cohesion
element together in the linear elastic range [34]. For pure mode I, II and III loading, after the
interfacial normal or shear strengths, the stiffnesses are gradually reduced to zero.
By using the linear-elastic softening behavior law, shown in Figure 2.27, for Mode I, II and
III of solicitation and applying the equation 2.124, the critical energy release rates G
IC
, G
IIC

and G
IIIC
can be determined as follows:

u
IC
= ]
3
()u
3
8
max,3
0
(2. 125)

u
IIC
= ]
13
()u
2
8
max,2
0
(2. 126)

u
IIIC
= ]
23
()u
1
8
max,1
0
(2. 127)

The mechanical properties required to determine the de-cohesion elements response are the
initial stiffness, or penalty stiffness K
p
, the critical energy release rates G
IC
, G
IIC
and G
IIIC
and
the interlaminar tensile and shear strengths of the material
3, ult
,
13, ult
and
23, ult
.


2 Theoretical Background 54



Fig. 2.27: Linear elastic-linear constitutive equation. [34]
The interfacial linear-elastic softening constitutive equation in Figure 2.27, can be represented
as follows:
1. For
i
<
o,i
, the constitutive equation is given by:

n = _
K
p
u u
u K
p
u
u u K
p
_ 6 = D 6 (2. 128)

2. For
o,i
<
i
<
max,i
, the constitutive equation is given by:

o = (I -F)D6 (2. 129)

Where I is the identity matrix and E a diagonal matrix defining the position of the
integration point in the softening.
3. For
i

max,i
, the penalty stiffness decrease to zero. It is still however necessary to
avoid the interpenetration of the crack faces, as a result, the normal penalty stiffness is
re-applied when interpenetration is detected.
The re-application of the normal stiffness in case of interpenetration is a common procedure
to deal with contact issues. The value of the applied stiffness has to be carefully chosen,
sufficiently high to avoid the interpenetration of the crack faces but not too much to avoid
numerical problems.
2.4.3.3.3 Mixed-mode Delamination
In this case, the propagation of the delamination can start even before one of the limit
tractions for pure mode loading individually attained. The constitutive response of the
cohesive zone is therefore given as a combination of the constitutive equations of the single
modes, their combination being ruled by the mixed-mode propagation criteria used (see
section 2.4.3.1). Many mixed-mode delamination criteria can be defined [35]. A frequently
2 Theoretical Background 55


used method is to consider only the mode I and a generic shear mode, regrouping the modes II
and III, using the linear elastic-linear softening constitutive equation. The mixed-mode
behavior under these assumptions is illustrated in Figure 2.28.

Fig. 2.28: Mixed-mode behavior for linear elastic-linear constitutive equation. [34]
The shear relative displacement (
shear
) and the total relative displacement can be defined as:

o
shcu
= o
2
2
+o
3
2
o = mox(u, o
1
)
2
+o
shcu
2
(2. 130)

Where the subscripts 1, 2 and 3 designate the values associated with the models I, II and III
respectively.
The shear traction t
shear
and the total traction t are defined as:

t
shcu
= t
2
2
+t
3
2
t = t
1
2
+t
shcu
2
(2. 131)

Having defined all these values, the onset relative displacement, for the mixed-mode has to
be determined. It can be performed using a delamination mixed mode, as is shown in section
2.3.5, in order to determine the penalty stiffness K for the mixed-mode. With the purpose of
determining the softening in the second linear part of the constitutive law, a mixed-mode
propagation criteria from is used to determine
f
, as illustrates Figure 2.28. For this goal the
equation 2.120 can be used, where G
I
and G
shear
can be expressed as follows:

2 Theoretical Background 56


0
I
=
K6
1
c
6
1
]
2
0
II
= 0
III
= 0
shcu
=
K6
shccr
c
6
shccr
]
2
(2. 132)

This combination of the equations 2.130 and 2.132 with the damage evolution criteria enable
to determine
f
.
2.4.3.3.4 Comments on the Cohesive Zone Models
The interface elements may either be infinitely thin or having a finite thickness. The initial
stiffness of the cohesive zone high stiffness can lead to numerical difficulties, too low values,
can introduce additional compliance into the model. It also appears that the shape of the
interface constitutive law is not of an extreme importance, as long as the fracture toughness is
correctly consistent with the stiffness and strength of the material being modeled. It may
however influence the stability of the analysis.
Another sensitive parameter of such zones is the mesh size. The mesh must be sufficiently
fine to ensure that enough interface elements exit within the cohesive zone length at the point
of crack propagation. However, the convergence of the problem tends to get more difficult
with a finer mesh, so especially for large structures, the right mesh density has to be adopted.
A general recommendation for the over meshing of the de-cohesion element is to be about
five times denser by the inclusion of these elements is also a reason why dynamic analyses,
avoiding by this way the difficulty of having to obtain a converged solution at each step.

3 FE Modeling of the ARCAN Test 57


3 FE Modeling of the ARCAN Test
This chapter describes the development of an ARCAN test procedure that permits to
characterize the fiber and matrix behavior in coupled I/II opening modes, by the loading of
fiber reinforced carbon composite bolted joint specimens to pure shear, pure tension and
combined shear-tension loads. The main advantage of the method is to apply, with single
specimen geometry, different tangential/normal loading ratio in order to fully identify the
delamination envelope, from pure shear to pure tension stress.
3.1 Description of the ARCAN Physical Test
The final objective is to be able to take into account complex failure mechanism including
delamination in the numerical model, under the application of shear-tension combined loading
as a virtual test and thus to improve the prediction capabilities of the developed FE model to
predict the failure behavior of bolted joints in composite structures for aerospace applications.
The considered experimental and numerical activities rely on the development and evaluation
of an ARCAN test procedure that generates controlled mixed shear/tension loading, using
single specimen geometry.
3.1.1 Geometry
The ARCAN Test Fixture consists of a set of different parts that may be described as follows:
the specimen, the test bolt and nut, retaining blocks and bolts, the semi-circumferential parts,
the load cell, etc. For the purpose of numerical model development, it is decided to approach
on focusing on the core parts first, taking into account as much details of the test rig into
consideration, as necessary.
3.1.1.1 Specimen
The specimen is composed of two laminates of quasi-isotropic carbon composite, as shown in
Figure 3.1, where L is the length, W the width, T the thickness of the laminate, D
1
the
diameter of the hole where the test bolt is assembled, D
2
the diameter of the hole for the
retaining bolts, C the distance from an edge of the laminate to the center of the closest hole in
the direction of the width and E the distance from an edge of the laminate to the center of the
closest hole in the direction of the length.

3 FE Modeling of the ARCAN Test 58



Fig. 3.1: Drawing of the Composite Coupon
3.1.1.2 Test Fastener
The composite coupons are fastened using a bolt and a nut. The bolts used for this purpose are
prEN6115-04 T4-7 and prEN6114-04 T4-7, the first type is a protruding-head bolt and the
second type is a countersunk-head bolt, both are made of titanium and two different diameter
codes has been used, 4, which means a shank diameter of 6.35 mm, and 5, which means a
shank diameter of 7.94 mm. The nuts used are a self locking nut, which has the designation
ASNA2536. The particularity of such a kind of nut is that under the pre-torque, it gets
tightened and deformed that entails its locking. Drawings of the nut and the bolt models are
given in Figure 3.2.
3 FE Modeling of the ARCAN Test 59



Fig. 3.2: Drawing of the Protruding-Head Bolt
3.1.1.3 Test Fixture
The test fixture is composed of the parts, four semi-circumferential parts, two retaining blocks
and eight counter-sunk head fasteners, the main function of which is to transfer the load from
the load cell to the joint under test, as shown in Figure 3.3.

Fig. 3.3: Drawing of the Test Fixture
3.1.1.4 Retaining Blocks
The Retaining Blocks (shown in Figure 3.4) are two blocks made of steel, with four holes on
the side, in order to be joined with four pins to the semi-circumferential parts. It has also four
holes on the squares going from the top to the bottom, where the fasteners which transfer the
load to the coupon are assembled, and a large hole in the middle, in order to avoid interaction
with the head of the test bolt, as well as with the test nut.
3 FE Modeling of the ARCAN Test 60



Fig. 3.4: Drawing of the Retaining Block
3.1.1.5 Fasteners
The composite coupons are joined to the retaining blocks, each composite coupon with four
bolts and nuts to retaining block respectively. The bolts used for this purpose are prEN6114-
04 T7-30, which are countersunk-head bolts made of titanium, having a shank-diameter of
11.11mm. The nuts used are a self locking nut, which has the designation ASNA2536. The
particularity of such a kind of nut is that under the pre-torque, it gets tightened and deformed,
that entails its locking. Drawings of the nut and the bolt models are given in Figure 3.5.

Fig. 3.5: Drawing of the Countersunk-Head Bolt
3.1.1.6 Semi-Circumferential Parts
The semi-circumferential parts are made of a very rigid alloy of steel. These parts are very
important for this test, because they are responsible for the possible change in the loading
angle, for which reason this particular test is one of the most important tests in the analysis of
composite bolted joints, as shown in Figure 3.6.
3 FE Modeling of the ARCAN Test 61



Fig. 3.6: Drawing of the Semi-Circumferential Part
3.1.2 Test Description
The ARCAN test (see Figure 3.7) is used to apply combined tension and shear loads on
specimens. It test rig is made up of 2 separate pairs of almost half circular symmetrical
metallic components. Each of them presents regularly opposite distributed load-application
points every 15 and are joined to a rigid steel retaining block. The composite specimen is
rigidly fixed on both retaining blocks through 4 counter sunk bolts each.
Fig. 3.7: ARCAN Test under 45 loading (left) and under 90 loading (right)
3.1.3 Test Procedure
Once the ARCAN test rig is set up on a static hydraulic machine and the force is applied, the
load transfers from one metallic rigid segment to the other through the specimen which
constitutes the weak point of the assembly. The rotation of the half circular metallic segments
and attached specimen with respect to the hydraulic machine load axis generates different
shear-tension loading ratios on the composite specimen surface. The average tension and
shear loads are functions of the load angle and the force F (eq. (3.1) and (3.2)).
3 FE Modeling of the ARCAN Test 62


Provided that the specimen dimensions are appropriate, one can assume that the shear stress
field is uniform at the mid-section in pure shear load configuration ( = 90).
F3= F cos (o) (3.1)
F1 = F sin (o) (3. 2)
3.2 Description of the ARCAN FE Virtual Test
3.2.1 Development of a Tool for a Parametric Detailed FE
Model
In the following paragraphs, a description of the plugin-tool will be presented, that is
developed to build up an automated and parametric detailed Finite Element Model of the
composite bolted joint in an ARCAN Test; that aims to also include partially, a few
components of the test rig itself. This tool is developed with the scripting language Python.
The main rationale behind this development is to be able to perform parametric studies on a
range of geometrical and physical parameters in simple specimen level tests, with no or very
slight modification in the script.
In the following section, the tool will be discussed in detail, first in the form of a simple
flowchart depicting the main idea, and then a step-by-step user-interface module of the tool.
3.2.1.1 Python-Script Flowchart
The internal performance of the simulator consists in a main script, so called Terminal,
which is itself, in charge of script and data management. This script receives the inputs from
the User-Interface Modules and manages all the sub-scripts sending back to the user the
Output requested, that is the 3D model. As presented in Table 5, the tool is composed of the
main script and six sub-scripts; each of these sub-scripts is in charge of one of the main parts
of the FE Model automatically generated with the Abaqus CAE Pre-processor. Some of these
scripts are also divided into several functions in order to harmonize the code.
3 FE Modeling of the ARCAN Test 63


Tab. 5: Tasks distribution between scripts

3.2.1.2 User-Interface Modules
At a first step, when the plugin-tool is called into the Pre-processor by the end-user, the
following graphical user-interface appears, that allows the selection of several sub-menus.
Under each sub-menu, a range of model and property definitions appears, that allows the
selection of the required properties to create the model with the relevant inputs, and simulate
the generated model.

Fig. 3.8: User-Interface Test Data Selection
Terminal
PartCreator
Retainig
Block
Bolts
Coupon
Nuts
Property
Materials
Sections
Assembly
Meshing
Interaction
Steps
3 FE Modeling of the ARCAN Test 64


1. The first sub-menu, Test Data, allows the user to select at a first step, the bolt head, type
of nut, diameter code or the approach for the model, as presented in Figure 3.8.

Fig. 3.9: User-Interface Coupon Data Selection
2. In the second sub-menu, the user can select the main coupon geometry parameters that
need to be simulated, as shown in Figure 3.9.

Fig. 3.10: User-Interface Pretension/Displacement Selection
3. At a third step, under the sub-menu Pretension/Displacement Data, the user can select the
main test parameters that he wants to simulate, such as the values for the pretension of the
fasteners and loading angle applied, as presented in Figure 3.10.
3 FE Modeling of the ARCAN Test 65



Fig. 3.11: User-Interface Abaqus Version/Input File
4. At a final step, under the sub-menu Abaqus Version/Input File, a compatibility mode has
been developed for solutions to be run in Abaqus version 6.10-1 and also for previous
versions. The main reason behind this is to solve possible incongruence within the code
between versions of the tool. The user can also select the automatic creation of the input
file (see Figure 3.11).
Once the needed parameters for the simulation have all been selected, the user can press the
button OK that allows the user-interface to submit these values to the terminal, which then
starts to manage the scripts and the model is automatically generated. The time required by
the simulator to generate the whole model is around 2 minutes.
3.2.2 Geometry
As is shown in Figure 3.3, the specimen used is composed of different parts, a pair of
coupons, the test bolt and nut, holding blocks and retaining bolts, which are all well
reproduced in the simulation.
3.2.2.1 Composite Plate
For the purpose of modeling, the composite coupons are divided in three different parts, first
of all, the coupon core, which is the surrounding area of the test bolt and the most important
part for the study, it is very finely meshed in order to build up better numerical approach to
capture the physical behavior of the joint. A second part would be the coupon outer cores,
which represent the area surrounding the retaining bolts, in order to capture possible small
cracks or bearing initiation around the holes that could affect in the test results, these parts are
also finely meshed, but not as fine as the coupon core. Both coupon core and coupon outer
cores were built up applying the said numerical approaches. The last part of the coupon is the
core vicinity; this part surrounds all the cores and it has a coarser mesh and an isotropic
3 FE Modeling of the ARCAN Test 66


material approach in order to save computational time and overall gets a general overview of
the behavior as no failure is expected in this area.
3.2.2.2 Fastener
For the test joint, two kind of bolt have been tested: countersunk head (prEN6114-04) and
protruding head (prEN6115-04). The test bolt is cut into two parts: the head and a further pin
which are connected by a connector. The connector is loaded for pretension and joins the
mating surfaces of the head and pin by distributed coupling that allows rotation and
translation in the mating surface plane. The nut used for the test fastener has been
ANSA2536.
Fig. 3.12: Nut (left) and Protruding-Head Bolt (right) meshed
3.2.2.3 Test Fixture
Two kinds of fasteners has been used for the test and also for the simulation, for the retaining
bolts as it is said in the norm, prEN6114-04 T7-30 is used (see Figure 3.8)
3.2.2.3.1 Retaining Blocks
The holding blocks have been well reproduced as a block made of steel with holes for the
retaining bolts and for a further implementation of the semi-circumferential blocks and a big
hole in the middle (see Figure 3.9).
3 FE Modeling of the ARCAN Test 67



Fig. 3.13: Retaining Block meshed
3.2.2.3.2 Retaining Fasteners
For the retaining fasteners, it has been used countersunk head (prEN6114-04). As for the test
bolt, the retaining bolt is cut into two parts: the head and a further pin which are connected by
a connector. The connector is loaded for pretension and joins the mating surfaces of the head
and pin by distributed coupling that allows rotation and translation in the mating surface
plane. The nuts used for the retaining fastener is ANSA2536, the same type of nut used for
the test joint, but with a greater diameter code.


Fig. 3.14: Countersunk-Head Bolt (left) and Nut (right) meshed
3 FE Modeling of the ARCAN Test 68


3.2.3 Meshing and Element Properties
For all parts solid elements have been used continuum 3 dimensional elements with reduced
integration points (C3D8R) and cohesive 3-dimentional elements (COH3D8) for the approach
of cohesive elements, as shown in Table 6.
Tab. 6: Applied Element Types
Test Bolt Solids C3D8R
Steel Nut Solids C3D8R
Retaining Block Bolts Solids C3D8R
CFRP Coupons Solids C3D8R
Resin Layers (if applied) Cohesives COH3D8

The mesh sizes were varied within the various simulations considerably, especially within the
CFRP coupon core. Therefore the mesh of the other parts will be presented first.
3.2.3.1 Test Fastener
The test bolt is divided into the head and further a pin, which are connected by a connector.
The connector is loaded for pretension and connected to the mating surfaces of the head and
the pin by distributed coupling, which remains rotation and displacement in the mating
surface plane. To maintain a good level of accuracy the mesh density in the vicinity of the
holes and their surrounding parts, namely the bolts and the nuts have to be the same.
Tab. 7: Meshing and Element Controls of Central Nut
Meshing Technique Sweep Advancing Front
Global element size 0.5
Hourglassing Enhanced
Element Deletion Yes

3 FE Modeling of the ARCAN Test 69


Tab. 8: Meshing and Element Controls of Central Bolt PAN-Head
Meshing Technique Sweep Advancing Front
Global element size 0.5
Hourglassing Enhanced
Element Deletion Yes
3.2.3.2 Core Vicinity
The Core-Vicinity is less influenced by the tension peaks around the fastener holes. Therefore
a coarse mesh is tolerable.
Tab. 9: Meshing and Element Controls of the Core Vicinity
Meshing Technique Structured
Global element size 3.0
Hourglassing Enhanced
Element Deletion Yes

Fig. 3.15: Core Vicinity meshed
3 FE Modeling of the ARCAN Test 70


3.2.3.3 Coupon Core
The Coupon Core suffers from high tension peaks around the central bolt hole. Furthermore a
high mesh density is necessary for a proper damage modeling so that the time between loss of
contact due to element deletion and regaining contact is minimized. To meet these needs
which especially apply to the close bolt hole vicinity and, at the same time, to save
computational cost, a variable element size in decreasing order towards the bolt hole edge is
applied to the Coupon Core.
Tab. 10: Meshing and Element Controls of the Coupon Core
Meshing Technique Structured
Global element size 0.6
Hourglassing Enhanced
Element Deletion Yes


Fig. 3.16: Coupon Core meshed
3.2.3.4 Coupon Outer Cores
The Outer Cores, being in contact with the retaining bolts, need to be meshed finer than the
Core-Vicinity. On the other hand, they are less loaded than the central bolt hole in the Coupon
Core and damage in these elements is not expected.
3 FE Modeling of the ARCAN Test 71


Tab. 11: Meshing and Element Controls of the Coupon Outer Cores








Fig. 3.17: Coupon Outer Core meshed
3.2.3.5 Cohesives (Approach of Cohesive Elements)
If resin layers were applied using layers of Cohesive Elements between the Cores Plies, the
element size was chosen so that along every edge of an adjoining continuum element three
Cohesive Elements were located. The layer thickness was chosen to be 3% of the Ply
thickness.

Meshing Technique Structured
Global element size 0.6
Hourglassing Enhanced
Element Deletion Yes
3 FE Modeling of the ARCAN Test 72


Tab. 12: Meshing and Element Controls of Retaining Block
Meshing Technique Sweep Medial Axis
Global element size 0.3
Hourglassing Enhanced
Element Deletion Yes
3.2.3.6 Retaining Block
The Retaining Blocks have been designed as in reality with four transversally holes to be
joined to the semi-circumferential parts through four rods. For this simulation the internal
surfaces of these holes have been taken as application point of the load, in order to get more
realistic results. Due to the complexity of the geometry, two kinds of mash have been used
with the aim of avoiding analytical problems during the analysis.
Tab. 13: Meshing and Element Controls of Retaining Block
Meshing Technique Sweep Medial Axis (Vicinity of the Retaining Bolts)
Structured (External area)
Global element size 3.0
Hourglassing Enhanced
Element Deletion Yes
3.2.3.7 Retaining Fasteners
The same procedure has been used as for the Test Bolt, the retaining bolts are divided into the
Head and a further Pin which are connected by a Connector. The Connector is loaded for
pretension and connected to the mating surfaces of the Head and Pin by Distributed Coupling
which remains rotation and displacement in the mating surface plane. To maintain a good
level of accuracy the mesh density in the vicinity of the holes and their surrounding parts,
namely the bolts and the nuts have to be the same.
3 FE Modeling of the ARCAN Test 73


Tab. 14: Meshing and Element Controls of the Retaining Bolts
Meshing Technique Sweep Advancing Front
Global element size 1.0
Hourglassing Enhanced
Element Deletion Yes
Tab. 15: Meshing and Element Controls of the Retaining Nuts
Meshing Technique Sweep Advancing Front
Global element size 1.0
Hourglassing Enhanced
Element Deletion Yes
3.2.3.8 Cohesive Elements
There are several ways of modeling the inter-laminar damage within a composite material (as
shown in 2.4.3). Cohesive elements can be considered as being composed of two faces with
infinitely small thickness, as represented in Figure 3.14. The relative motion between the two
faces measured along the direction of the thickness represents the opening of the surface.
Stretching and shearing of the mid-surface of the element are associated with membrane
strains in the cohesive element. It is however assumed that the cohesive elements do not
generate any stresses in a purely membrane response [21].

Fig. 3.18: Spatial representation of a three-dimensional cohesive element (left) and deformation modes of
a cohesive element (right). [21]
Very thin layers (around 3% of the solid layer thickness, thus 0.0075 mm) have been added in
the lay-up. It has been finely meshed as recommended by Harper et al. [34]. The average size
of the mesh is 0.3mm, four times the density as recommended by Diehl [32]. There was an
effort to lower the value for the mesh size of the cohesive elements but due to the capacity of
the cluster, the previous value was used, with also a single element in the thickness. While
3 FE Modeling of the ARCAN Test 74


using cohesive elements, it has to be taken into consideration the important issues that are
specific to these elements. Such issues include special considerations associated with using
cohesive elements in conjunction with contact interactions, potential degradation of the stable
time increment size in Abaqus/Explicit. In three-dimensional problems, the traction-
separation-based model assumes three components of separationone normal to the interface
and two parallel to it; and the corresponding stress components are assumed to be active at a
material point. In two-dimensional problems, the traction-separation-based model assumes
two components of separationone normal to the interface and the other parallel to it; and the
corresponding stress components are assumed to be active at a material point. The traction-
separation constitutive response option (see section 2.4.3) is used. These cohesive sections
have been assigned a small transverse shear stiffness that is supposed to provide additional
stability to the cohesive elements. The critical energy releases rates come from calibration
tests. The other data have been determined using the recommendation of Diehl [32]. The
failure criteria used is the maximum stress failure criteria and the damage evolution is
predicted using the Benzeggagh-Kenane fracture criterion. Damage is assumed to initiate
when the maximum nominal stress ratio (as defined in the expression below) reaches a value
of one. This criterion can be represented as:

mox _
t
IC

t
IC
c
,
t
IIC
t
IIC
c
,
t
IIIC
t
IIIC
c
_ = 1 (3. 3)

Where t
IC
, t
IIC
and t
IIIC
represent the normal (along the local 3-direction) and the two shear
tractions (along the local 1- and 2-directions in three dimensions) respectively and t
IC
o
,
t
IC
o
and t
IC
o
represent the peak values of the nominal stress when the deformation is either
purely normal to the interface or purely in the first or the second shear direction, respectively.
The symbol used in the discussion represents the Macaulay bracket with the usual
interpretation. The Macaulay brackets are used to signify that a pure compressive deformation
or stress state does not initiate damage.
It is a mixed mode failure criteria adapted when the critical fracture energies during
deformation, purely along the first and second shear directions, are the same (0
IIC
= 0
IIIC
). It
is given by:

0
C
= 0
IC
+(0
IIC
-0
IC
) [
u
II
+u
III
u
I
+u
II
+u
III

u
(3. 4)

Where the power is a material parameter.
From calibration tests we obtain G
IC
, G
IIC
and G
IIIC
,

t
IC
(t
IIC
, t
IIIC
) = 2
u
IC
(u
IIC
,u
IIIC
)
6
rctic
6
]cilurc
2
(3. 5)
3 FE Modeling of the ARCAN Test 75


*Material, name=Cohesive
*Density
1.2000e-9
*Elastic, type=traction
6234.00, 21451.11, 21451.11
*damage initiation, criterion=MAXS
46.755, 160.883, 160.883
*damage evolution, type=energy, mixed mode behavior=BK,
softening=exponential, mode mix ratio=energy, power=1.5
0.28053, 0.9653, 0.9653

The constraints used to tie the cohesive layers with the rest of the model are presented
hereafter.
3.2.4 Constraints & Interactions
3.2.4.1 Tie Constraint
A Tie Constraint ties two separated surfaces together so that there is no relative motion
between them. Such constraint has been used to bind each surface of each CFRP layer to each
of its adjacent layers for the initial approach without cohesives and to each of its adjacent
cohesive layers for the approach with cohesive elements. It has to be stated that the Tie
Constraint is also used to attach the cores, where the numerical and material approaches are
applied, with the core vicinity level by level. And finally the tie constraint has been applied to
secure the nuts to the pins for each of the nine fasteners in order to reproduce the behavior of
the joint in the reality.
3.2.4.2 General Contact
General Contact has been used to establish all the contacts within the assembled structure.
The general contact algorithm also enforces contact between Eulerian materials and
Lagrangian surfaces. This algorithm automatically compensates for mesh size discrepancies to
prevent penetration of Eulerian material through the Lagrangian surface. The all-inclusive
surface that is defined by Abaqus/Explicit can be used to enforce contact between all Eulerian
materials and all Lagrangian bodies in a model. Eulerian-Lagrangian contact is enforced only
for Lagrangian surfaces defined on solid and shell elements. Other surface types, such as
beam edges and analytical rigid surfaces, are ignored. [21]
3 FE Modeling of the ARCAN Test 76



Fig. 3.19: Topology of an eroding contact surface. [21]
General Contact has been used to render the contact between the different parts. Three fiction
coefficients have been applied, 0.3 for all the contacts between metal and metal, as well as
between metal and composite, and 0.5 for the contacts between composite and composite. For
the cohesive elements approach, as the surfaces of cohesive layers have been tied to their
adjacent solid layers, they do not need to be included in any contact definition. Nevertheless,
these layers shoud be included in the contact definition to avoid the penetration of the layers
whenever some cohesive elements would get damaged and deleted.
*Contact, op=NEW
*Contact Inclusions, ALL EXTERIOR
*Contact Inclusions
,surf_interior
surf_interior,
*Contact property assignment
, , metalcfkIntProp
contact_core_head,contact_core_tail, cfkcfkIntProp
*Surface property assignment, property=FEATURE EDGE CRITERIA
, NO FEATURE EDGES
*contact controls assignment, type = scale penalty
contact_core_head,contact_core_tail, 2.0
headCoupon_boltHead, boltHead_headCoupon, 2.0
contact_coupon_to_asna, contact_asna, 2.0
bolthead_surf_no_contact, tailCore-2_surf_hole, 2.0
bolthead_surf_no_contact, tailCore-1_surf_hole, 2.0

3 FE Modeling of the ARCAN Test 77


3.2.4.3 Cohesive Behavior

Fig. 3.20: Retaining Fastener Connector
The failure criteria used is the maximum stress failure criteria and the damage evolution is
predicted using the Benzeggagh-Kenane fracture criterion. Damage is assumed to initiate
when the maximum contact stress ratio (as defined in the expression) reaches a value of one.
This criterion can be represented as

mox _
t
IC

t
IC
c
,
t
IIC
t
IIC
c
,
t
IIIC
t
IIIC
c
_ = 1 (3. 6)

Where t
IC
, t
IIC
and t
IIIC
represent the normal (along the local 3-direction) and the two shear
tractions (along the local 1- and 2-directions in three dimensions) respectively and t
IC
o
,
t
IC
o
and t
IC
o
represent the peak values of the contact stress when the separation is either purely
normal to the interface or purely in the first or the second shear direction, respectively
The damage evolution law describes the rate at which the cohesive stiffness is degraded once
the corresponding initiation criterion is reached. Conceptually, similar ideas apply for
describing damage evolution in cohesive surfaces. A scalar damage variable, D, represents the
overall damage at the contact point. It initially has a value of 0. If damage evolution is
modeled, D monotonically evolves from 0 to 1 upon further loading after the initiation of
damage. The contact stress components are affected by the damage according to:

t = _
(1 - )t
n

t
n

1
t
n


(3. 7)

t
s
= (1 - )t
s
(3. 8)

t
t
= (1 - )t
t
(3. 9)
3 FE Modeling of the ARCAN Test 78


Where t
n

, t
s

, and t
t

are the contact stress components predicted by the elastic traction-


separation behavior for the current separations without damage. To describe the evolution of
damage under a combination of normal and shear separations across the interface, it is useful
to introduce an effective separation [33] defined as:

o
m
=
n

2
+
t
2
+
s
2
(3. 10)

It is a mixed-mode failure criteria adapted when the critical fracture energies during
deformation, purely along the first and second shear directions, are the same (0
IIC
= 0
IIIC
), as
shown in section 3.2.2.8.
*Elastic, type=traction
6234.00, 21451.11, 21451.11
*damage initiation, criterion=MAXS
46.755, 160.883, 160.883
*damage evolution, type=energy, mixed mode behavior=BK,
softening=exponential, mode mix ratio=energy, power=1.5
0.28053, 0.9653, 0.9653
3.2.5 User Subroutine
To define mechanical constitutive behavior of the CFRP in the solids elements, the model
considered is based on the built-in damage model for CFRP available in both
Abaqus/Standard and Abaqus/Explicit. The damage in the CFRP is simulated using the model
proposed by Linde et al. [36], which is implemented in the user subroutine, called in
Abaqus/Explicit as VUMAT and is discussed below.
In the VUMAT model, the damage initiation criteria is expressed in terms of strains. Unlike
the built-in model in Abaqus, which uses four internal (damage) variables, the VUMAT
model uses two damage variables to describe damage in the fiber and matrix without
distinguishing between tension and compression. Although the performance of the two
models is expected to be similar for monotonic loads, such as in this example problem, the
results obtained might differ considerably for more complex loads in which, for example,
tension is followed by compression. For the VUMAT model, if the material is subjected to
tensile stresses that are large enough to cause partial or full damage (the damage variable
corresponding to this damage mode will be greater than zero), both tensile and compressive
responses of the material will be affected. However, in the case of the built-in damage model,
only the tensile response will be degraded while the material compressive response will not be
affected. In many cases the latter behavior is more suitable for modeling fiber-reinforced
composites. In this section the governing equations for damage initiation and evolution as
proposed by Linde et al. (2004) are discussed, followed by a description of the user
subroutine VUMAT implementation [19].
3 FE Modeling of the ARCAN Test 79


Damage in the fiber is initiated when the following criterion is reached:

]
= _
s
L
],t
s
L
],c
e
L
2
+_e
L
],t
-
[s
L
],t

2
s
L
],c
_e
L
(3. 11)

]
> e
L
],t
(3. 12)

Where e
L
],t
=
c
L
],t
C
11
, e
I
,c
=
o
I
,c
C
11
and C
ij
are the components of the elasticity matrix in the
undamaged state, being o
L
],t
and o
L
],c
the stress limits for traction and compression
respectively. Once the above criterion is satisfied, the fiber damage variable, d
f
, evolves
according to the equation:

J
]
= 1 -
s
L
],t
]
]
c
_
-C
11
s
L
],t
_]
]
-s
L
],t
]L
C
G
]
_
(3. 13)

Where I
C
is the characteristic length associated with the material point. Similarly, damage
initiation in the matrix is governed by the criterion:

m
= _
s
T
],t
s
T
],c
e
1
2
+_e
1
],t
-
[s
T
],t

2
s
T
],c
_e
1
+
s
T
],t
s
LT
]
e
L1
2
(3. 14)

m
> e
1
],t
(3. 15)

Where e
1
],t
=
c
T
],t
C
22
, e
T
],c
=
c
T
],c
C
22
and e
LT
]
=
:
LT
]
C
44
being o
L
],t
, o
L
],c
and
L1
]
the stress transversally
limits for traction and compression and the shear stress limit respectively. The evolution law
of the matrix damage variable, d
m
is:

J
m
= 1 -
s
T
],t
]
m
c
_-
C
22
s
T
],t
_]
m
-s
T
],t
]L
C
G
m
_
(3. 16)

3 FE Modeling of the ARCAN Test 80


During progressive damage the effective elasticity matrix is reduced by the two damage
variables d
f
and d
m
, as follows:

C
d
=
l
l
l
l
l
l
l
(1 -J
]
)C
11
(1 -J
]
)(1 -J
m
)C
12
(1 -J
]
)C
13
u u u
(1 -J
]
)(1 -J
m
)C
12
(1 -J
m
)C
22
(1 -J
m
)C
23
u u u
(1 -J
]
)C
13
(1 -J
m
)C
23
C
33
u u u
u u u (1 -J
]
)(1 -J
m
)C
44
u u
u u u u C
55
u
u u u u u C
66
1
1
1
1
1
1
1
(3. 17)
*Material, name=CFK-Mat
*Density
1.588e-09,
*Depvar, delete=9
10,
*User Material, constants=14
135000.0, 8500.0, 4200.0, 3150.0, 0.35, 0.35,
3164.0, 1172.0
110.693, 260.0, 143.482, 0.63, 141.73, 3.0

Fortran VUMAT properties:
props(1)= E1 young's modulus in 1 direction
props(2)= E2 young's modulus in 2 direction
props(3)= g12 shear modulus in 1-2 plane
props(4)= g23 shear modulus in 2-3 plane
props(5)= xnu12 poisson ratio poi_12
props(6)= xnu23 poisson ratio poi_23
props(7)= sigu1t ultimate stress in 1 direction in tension
props(8)= sigu1c ultimate stress in 1 direction in
compression
props(9)= sigu2t ultimate stress in 2 direction in tension
props(10)= sigu2c ultimate stress in 2 direction in
compression
props(11)= sigu12 ultimate shear stress
props(12)= gmat fracture energy in matrix
props(13)= gfib fracture in fiber
props(14)= eleDel element deletion criteria

3 FE Modeling of the ARCAN Test 81


3.2.6 Loading Steps
3.2.6.1 Initial Pre-tension
In order to be closer to the experimental test, the pretension in the bolts has been implemented
and its influence on the result examined. Because of the fact that Abaqus Explicit does not
have an in-built method to model pretension in mechanically fastened joints, connector
elements will be used.

Fig. 3.21: Test Fastener Connector
Connector elements are elements which help to simulate the interactions between parts of an
FE model. The procedure used to model pretension in Abaqus Explicit by using connectors
can be summarized as follows:
1. The bolt is cut into two parts as presented in Figure 3.21. One part should include the
head of the bolt (Part 1), and the other one, the rest of the pin, which is in contact with
the nut (Part 2). It is very important to have the cross-section in an area of minimum
stress to avoid the early failure of the connector element.
2. The properties of both surfaces resulting from the sectioning of the bolt have to be
assigned respectively to a reference point.
3. A connector element with the properties Cardan and Cartesian (shown in Fig.) can be
built between the parts by selecting the two reference points through a wire.
4. Finally, a connector load is assigned to the connector, 5500.0N for the test bolt and
13501.35N for the retaining block bolts. It is important to know that the application of
the connector force only depends on the local connector coordinate system and the
order in which the reference points defining the connector were chosen. Because the
connector force is applied on the second selected reference point, attention should be
paid in the orientation of the local connector coordinate system in order to determine
the sign convention of the connector force.
3 FE Modeling of the ARCAN Test 82


3.2.6.2 Final Test Displacement
With the aim of reproducing the test displacement, the inner surfaces of the retaining block
are coupled to a representative point on each block where the displacement is applied, this
approach has been taken assuming that the load goes directly from the load cell to the coupon
through the test fixture, which on a first approach has been modeled up to the retaining blocks
only in order to simplify the simulation. To reduce the computational time, the displacement
has been applied to both representative points, not only on the upper block, keeping the lower
block constrained in all its degrees of freedom. The displacement has been applied with a
velocity of approximately 100 mm/min, in order to have a quasi-static analysis, using a
smooth step to avoid abrupt amounts of energy.


4 Analysis of Experimental Results 83


4 Analysis of Experimental Results
In this chapter, experimental results will be presented, analyzed and compared, beginning
with pure tension and pure shear test with protruding head bolts, the same test that has been
simulated in the frame of this work. This is then continued with a comparison of a set of
results tested under varying loading angles, in order to show the influence of this parameter
on the stiffness, damage initiation and evolution, first for a set of tests with a protruding head
bolt and then for a set with a countersunk head bolt. In both cases, the failure mode angle by
angle is discussed. This is then concluded with a comparison between the test results with a
protruding and a countersunk head bolt, in terms of change in maximum peak loads with
loading angle.
4.1 Protruding Head Bolt: Pure Tension

Fig. 4.1: Pure Tension Experimental Result with Protruding Head Bolt
It can be seen that the stiffness in the pure tension test is dominated by the bolt stiffness, due
to this fact, the fastener behaves elastically until the first initiation of crack at around 2.0 mm
of displacement, then the stiffness continues to grow and cracks appear to propagate.
0 0,5 1 1,5 2 2,5 3 3,5 4
R
F
[
k
N
]
U(mm)
Results90
4 Analysis of Experimental Results 84



Fig. 4.2: Experimental Result of Pure Tension Test

4.2 Protruding Head Bolt: Pure Shear

Fig. 4.3: Pure Shear Experimental Result with Protruding Head Bolt
It can be seen that the stiffness is much higher than in the previous case. In the pure shear test,
the stiffness is dominated mainly by laminate stiffness. At around 0.4mm displacement,
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
R
F
[
k
N
]
U(mm)
Results0
4 Analysis of Experimental Results 85


bearing starts to initiate, this leads to a drop in the stiffness at around 1.0mm, which is caused
by the fracture of the bolt head leading to bolt failure.


Fig. 4.4: Experimental Result of Pure Shear Test

4.3 Comparison of Tests with Protruding Head Bolt

Fig. 4.5: Comparison of the Experimental Results in Terms of Loading Angle for Protruding Head Bolt
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5
R
F
[
k
N
]
U[mm]
Results0 Results15
Results30 Results45
Results60 Results75
Results90
4 Analysis of Experimental Results 86


The set of tests used here is with a protruding head bolt with a diameter of 6.35mm, joined to
two 6.00mm thick composite coupons. It is to mentioned here that 0 (failure shown in Figure
4.2) means pure tension test and 90 (failure shown in Figure 4.3) pure shear test, in
accordance to the position of the semi-circumferential parts. For test at lower angles, there
was a large flexural component in the load versus deflection curve prior to cracks and
delamination within the coupons. At a test angle of 30 and 45 loading, a similar response
was seen, however, ultimately the protruding head fractured and instantaneously pulled
through. In some cases, white sparks were generated at the point of fracture; this presumably
coming from the titanium fastener as it passes through the composite. The maximum load
increased from an average of 14.98kN at 90 to 17.907kN at 0.
It can be appreciated that the elastic stiffness increases with the loading angle, fact that is
expected because E
L
>>E
T
, E
L
being the longitudinal Youngs modulus and E
T
the transversal
Youngs modulus. It can also be appreciated that the evolution of the failure mode is
influenced by the loading angle, starting at 0 with cracks in the composite under protruding
head prior to pull-through and laminate flex, at 30 with protruding head fracture and sparks
occurring leading to an instantaneous pull-through, bearing starts at around 45 and finally a
failure in the bolt occurs.
Tab. 16: Shear-Tension Failure Modes for Protruding Head Bolt
Loading Angle () Failure Mode
0 Laminate flex. Cracks in composite under protruding head prior to pull-through.
15 Laminate flex. Cracks in composite under protruding head prior to pull-through.
30 Laminate flex, pull-through cracking. Ultimately protruding head fracture with
sparks and instantaneous pull-through.
45 Bearing, pull-through cracking then protruding head fracture with instantaneous
pull-through.
60 Bearing then bolt failure at first thread.
75 Bearing, protruding head fracture then bolt failure at first thread.
90 Bearing, protruding head fracture then bolt failure at first thread.

4 Analysis of Experimental Results 87


4.4 Comparison of Tests with Countersunk Head
Bolt

Fig. 4.6: Comparison of the Experimental Results in Terms of Loading Angle for Countersunk Head Bolt
The set of tests used here is with a protruding head bolt of 6.35mm diameter, joined to two
6.00mm thick coupons. The damage was observed in the composite with the bolt remaining
intact until 60, when the bolt head begins to fracture. The maximum load increased from an
average of 15.374kN at 90 to 18.952kN at 0.
It can be appreciated that, as for the protruding head bolt, the elastic stiffness increases with
the loading angle, fact that is expected because E
L
>>E
T
. As the test angle increases, the
flexural component of the test decreases and the bearing component increases, hence the
shape of the load deflection curve changes. The initial slope becomes much steeper as the test
angle increases towards shear mode. The initial steep slope is due to shear on the fastener and
bearing within the coupon holes. As the load continues to rise, damage begins to be dominant
in the composite as the fastener tilts.
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
R
F
[
k
N
]
U(mm)
Results90
Results60
Results45
Results30
Results0
4 Analysis of Experimental Results 88


Tab. 17: Shear-Tension Failure Modes for Countersunk Head Bolt
Loading Angle () Failure Mode
0 Laminate flex. Cracks in composite under protruding head pre-pull-trough.
30 Laminate flex, pull-through cracking. Ultimately protruding head fracture with
sparks and instantaneous pull-through.
45 Fastener tilt, bearing and pull-through damage. Majority of damage in upper coupon.
60 Countersunk head fail, pull-through and bearing damage.
90 Fastener tilt, countersunk head fracture with bearing then pull-through damage.

4.5 Comparison in Terms of Maximum Peak Loads

Fig. 4.7: Comparison of the Experimental Results in Terms of Shear-Tension Maximum Load Peak
Each of the points in the diagram (Figure 4.5) represents the maximum peak load shared
between tension and shear load. It can be appreciated that the test using a protruding head bolt
reaches a higher load, showing an increment of around 2.63% (394N) in pure tension and
0
2
4
6
8
10
12
14
16
18
0 5 10 15 20
F
t

[
k
N
]
Fs [kN]
Countersunk Head Bolt
Protruding Head Bolt
4 Analysis of Experimental Results 89


around 5.84% (1040N) in pure shear and reaching a maximum increment near 45 around
11.86% (1891N).


4 Analysis of Experimental Results 90




5 Analysis of Simulation Results 91


5 Analysis of Simulation Results
As a numerical method for finding approximate solutions of partial differential equations by
using element discretisation techniques, the finite element method presents a great approach
for solving mechanical problems. Because of the flexibility in term of discretisation, modeling
and solving techniques, which influences the rapidity as well as the correctness of the
analysis, a compromise has to be found between efficiency and accuracy.
5.1 Solids without Cohesives
5.1.1 Pure Tension
Before starting with the parametric numerical studies, an approach between the test result and
the first finite element analysis model in a quasi-static test is realized. The FE model
presented in the previous chapter is the starting point and is taken as a reference model in all
subsequent comparison analyses. It has been computed without errors and the first
impressions are good. The kinetic energy (ALLKE) compared to the internal energy (ALLIE)
is relatively low and the total energy (ETOTAL) is near to 0 and remain constant throughout
the analysis, as required for a quasi-static analysis. In Figure 5.1, it can be seen that the
internal energy is plotted in blue, the kinetic energy in red and the total energy in green.

Fig. 5.1: Energy Plot for Simulation without Cohesives
5 Analysis of Simulation Results 92



Fig. 5.2: Simulation without Cohesives
In this first FE analysis, as a model without cohesive cannot reproduce inter-laminar failure
and delamination, the behavior of the force-displacement curve after the linear elastic domain
in comparison to the test is very different as represented in Figure 5.2.
There is a total absence of drop in the stiffness after the expected failure. The analysis
continues as purely elastic, not being able to capture the expected delamination failure in the
laminate, which means no element deletion as shown in Figure 5.3.
Nonetheless, as the failure mode is described as a laminate cracking under the bolt head, it
can be assumed that the failure originates from a damage in the 3-direction of the laminate.
But as the used VUMAT does not include damage in the 3-direction, further work is
necessary to develop a material model for capturing three-dimensional stress state in the
composite.
0 1 2 3 4 5 6
R
F
3
[
k
N
]
U3[mm]
SimulationwithoutCohesives
ExperimentalResult
5 Analysis of Simulation Results 93



Fig. 5.3: Deformed and Un-deformed Plot of the Central Core for Model without Cohesives (Pure
Tension)
5.1.2 Pure Shear
It has been computed without errors and the first impressions are good. The kinetic energy
(ALLKE) compared to the internal energy (ALLIE) is relatively low, nevertheless this
relationship is bigger than expected and a higher mass scaling is needed for further
improvements of the model. The total energy (ETOTAL) is near to 0 and remains constant
throughout the analysis as required for a quasi-static analysis. See in Figure 5.4, where the
internal energy is plotted in red, the kinetic energy in green and the total energy in purple.

5 Analysis of Simulation Results 94



Fig. 5.4: Energy Plot for Simulation without Cohesives

Fig. 5.5: Simulation without Cohesives
This simulation has been first performed for 2.5s only, which means for displacement of
2.5mm. This reduction has been performed in order to capture at a first step a good correlation
with the test result itself and the test was stopped at 2.3mm displacement. The aim of this
simulation is to reproduce reasonably well the test result, with the purpose of investigating the
complete failure behavior in a further study with a higher maturity level of 3D virtual testing.
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
R
F
1
[
k
N
]
U1[mm]
Solids_wocoh
ExperimentalResults
5 Analysis of Simulation Results 95


The correlation shown between the experimental and the simulation results for the stiffness
with the approach of solids without cohesive is impressively good for the stiffness as well as
for the failure mode. The instabilities reached between 1mm until the end of the simulation
may be caused by a small dynamic effect as shown in Figure 5.4 or due to the element
deletion criteria used, a further investigation that can lead to an interesting enhancement in the
model, in order to reproduce accurately the failure behavior.

Fig. 5.6: Deformed and Un-deformed Plot of the Central Core for Model without Cohesives (Pure Shear)

5.2 Cohesive Elements
5.2.1 Pure Tension
As previously mentioned, an approach between the test result and the first finite element
analysis model in a quasi-static test is realized. The FE reference model presented in the
previous chapter is taken as a starting point. It has been computed without errors and the first
impressions are good.
5 Analysis of Simulation Results 96



Fig. 5.7: Simulation with Cohesive Elements
The kinetic energy (ALLKE) compared to the internal energy (ALLIE) is relatively low and
the total energy (ETOTAL) is near to 0 and remain constant throughout the analysis as
required for a quasi-static analysis. See in Figure 5.7, where the internal energy is plotted in
red, the kinetic energy in green and the total energy in purple.

Fig. 5.8: Simulation with Cohesive Elements
0 1 2 3 4 5 6
R
F
3
[
k
N
]
U3[mm]
SimulationwithCohesiveElements
ExperimentalResults
5 Analysis of Simulation Results 97


It can be appreciated that the stiffness reached using this model is closer to the experimental
result as compared to the stiffness obtained using the reference model. It can also be seen that
at around 3mm the delamination failure begins, this occurs 1mm later that in the experimental
test. It is assumed that this is caused by the element deletion criteria within the user material
subroutine, which in a further study has been enhanced, and a better correlation is obtained, it
reaches a first drop representing first initiation of cracks and then it keeps increasing, this
presents several instabilities caused by the mesh element size which due to computational
capacity was reduced from 0.15mm, twice the recommendation of Diehl [32] to 0.3mm. In a
further investigation the element size for the cohesive elements can be reduced with the
purpose of getting a more stable result using a more mature hardware, increasing the RAM
memory at least to the double of the actual available memory.
It can be appreciated, as shown in Figure 5.9, that the failure in the composite occurs as
cracks and also some delamination failure in between the layers of composite.

Fig. 5.9: Deformed and Un-deformed Plot of the Central Core for Cohesive Elements Model (Pure
Tension)
5 Analysis of Simulation Results 98


5.3 Cohesive Contact
5.3.1 Pure Tension
It has been computed without errors and the first impressions are good. The kinetic energy
(ALLKE) compared to the internal energy (ALLIE) is relatively low and the total energy
(ETOTAL) is near to 0 and remain constant throughout the analysis as required for a quasi-
static analysis. See in Figure 5.10.

Fig. 5.10: Simulation with Cohesive Elements
In a first approach, it can be seen that the stiffness is very close to the stiffness of the
experimental result as expected using the approach of cohesive contact. It can be appreciated
that the simulation stopped at around 2.5mm due to excessive distortion in some of the solid
elements, this distortion as well as the big noise that appears is caused by a bug in Abaqus
6.10-1, a conflict between the general contact and the cohesive behavior that creates
incoherence in the numerical analysis. This problem has been solved for the version 6.11-1, so
due to the computational advantage of this method in comparison with the model with
cohesive solids it would be recommendable to run the simulation using newer versions of the
code in order to obtain a result with the benefits in terms of computational cost of the model
without cohesive and also the accuracy of the model using cohesive elements.
0 0,5 1 1,5 2 2,5 3 3,5 4
R
F
3
[
k
N
]
U3[mm]
SimulationwithCohesiveContact
ExperimentalResults
5 Analysis of Simulation Results 99



Fig. 5.11: Deformed and Un-deformed Plot of the Central Core for Cohesive Contact Model (Pure
Tension)
In Figure 5.11, the stress distribution around the test bolt hole in the moment previous to the
total failure of analytically coherence between the general contact and the cohesive behavior,
shown in Figure 5.12.
5 Analysis of Simulation Results 100



Fig. 5.12: Final Frame of the Central Core for Cohesive Contact Mode Showing Simulation Problem
5.4 Computational Cost
The simulations have been performed on an Intel High Performance Cluster with 8 CPUs per
node. Given the different features of the models, it is interesting to compare the computation
time for simulations performed with each on the same number of CPUs. The properties of the
three models and their characteristics regarding computational time without cohesive, with
cohesive elements and with cohesive contact performed using 64 CPUs, are shown in Table
18. Between the models without cohesive and the model with cohesive contact, there is no
significant difference in terms of computational cost, it can also appreciated that the number
of variables is the same, the only difference is the approach for joining the composite layers in
between. It can also be seen that mass scaling was necessary for all of the cases, due to the
high amount of variables and to minimize the dynamic effect. The addition of small elements
to capture the delamination failure behavior in the model using cohesive elements has a
considerable influence on the computational efficiency of the model itself. The elements
responsible of this sharp difference are really small (mesh size of 0.3mm and thickness of
0.0075mm) located between the layers of composite within the cores.
5 Analysis of Simulation Results 101


Tab. 18: Comparison of the Approaches in Terms of Computational Cost
Model Solid without Cohesives Solid with Cohesive
Elements
Solid with Cohesive
Contact
Number of Elements 909564 1903388 909564
Number of Nodes 1278695 3351207 1278695
Stable Time Increment (s) 5.0(x10-6) 5.0(x10-6) 5.0(x10-6)
Mass Scaling (%) 10 10 10
Computation Time (s) 5.1 5.1 5.1
CPU Time (h) 27.75 68.38 27.82

To get an idea on the dependency of the computational time in terms of CPU usage for the
models, the simulation without cohesives has been performed on different number of CPUs:
2, 4, 16, 32 and 64. It can be seen in Figure 6.1 that the reduction in time by using more CPUs
is decreasing. This evolution can be approximated by an exponential curve, as shown by the
tendency line in Figure 6.1. This effect comes from the time that the cluster needs to
communicate the values between the CPUs that are working in parallel, time that increases
with the CPU usage.

Fig. 5.13: Evolution of the computational time with the number of CPUs for the simulation with the
approach of Cohesive Elements
0
50
100
150
200
250
300
350
400
1 2 4 8 16 32 64
C
P
U

T
i
m
e

[
h
]
NumberofCPUs
5 Analysis of Simulation Results 102




6 Conclusion and Way Forward 103


6 Conclusion and Way Forward
Hereafter conclusions will be presented on the achievements of the simulation results as well
as suggestions of improvements for further work.
6.1 Conclusion
Nowadays the number of flights is growing exponentially because of which it is indispensable
to find new ways of saving fuel. Advanced composite materials present in an era of climate
change and rapidly decreasing fossil resources not only an opportunity for robust and
innovative weight saving structures for the aircraft industry, but also great challenges in
engineering as it can be observed in this study regarding the behavior of the ARCAN Test.
The manufacturing, maintenance and the need of reliable load carrying structures make
indispensable the use of mechanically fastened joints. Given the lack of experience with
composites in comparison to metal in general and with joints in particular, the design of
reliable composite structures is a matter of huge development as well as experimental costs.
In order to reduce these experimental costs and therefore reduce design lead time, advanced
numerical methods must be established to validate test results accurately. Based on this
assumption and by use of the finite element code Abaqus Explicit, a numerical and physical
parameters investigation on the behavior of a composite bolted joint under quasi-static
loading have been performed through the ARCAN Test.
About the simulation models the following remarks can be summarized.
For pure shear, a simple approach using solid elements for the layers of composite tied
together by tie contact offers a good approach due to the fact that the global stiffness of the
model is dominated mainly by laminate stiffness, in particular dominated by the stiffness in
direction 1 and 2, this behavior is well reproduced by the user subroutine.
For pure tension, the above-mentioned approach is good for the calculation of the stiffness,
but need to be enhanced with the purpose of capturing the delamination failure. For this
reason, the approach using cohesive elements has been implemented, showing even better
results for the stiffness and capturing the inter-laminar failure, the only problem that this
method has is the enormous computational cost in comparison with the previous approach.
For this reason, the approach of cohesive contact has been used. To obtain better results with
this approach, a more matured version of the FE code Abaqus Explicit is to be tested and
used, where such numerical incoherences during the analyses are said to have been removed.
6.2 Way Forward
Further analyses are needed in order to improve and optimize the FE and the physical test
correlation. Because a composite material in general and composite bolted joint in particular
is such a challenging field, better correlation between test and numerical specimens involves a
closer cooperation between manufacturing and testing as well development of enhanced and
advanced material models to characterize the composite material model behavior.
6 Conclusion and Way Forward 104


The User Material Subroutine used for this study does not take into account the difference
between tension and compression failure for the fiber neither for the matrix, for this reason it
is recommended to run the models using more complex User Subroutine, as the Enhanced
Hashin or the recently developed Larc05.
The approach with cohesive elements seems to be the most adequate for Abaqus 6.10-1, but
presents enormous computational cost. It can be used but in order to minimize this problem, it
is needed established a computational evolution in terms improvement in parallelization of
processors and number of CPU usage, as increasing the number of CPUs after a certain
amount that is not always lead to reduce computational cost.
For this reason, the use of Abaqus 6.12-1 is highly recommended to perform a successfully
cohesive contact approach avoiding the incoherence in the numerical analysis between
cohesive behavior and general contact.
Concluding with the way forward, it would be interesting to implement two different
approaches into the model; one approach for the initial pretension in order to simulate better
the behavior of the joint under and after pretension and another approach for multi-scale
approaches using multi-processors with the aim of distributing wiser the CPUs, using more
for fine meshes and less for coarse meshes.


A Apendices 105


A. Apendices
a. Data and Software-code
i. User Material Subroutine [21]
C 4X4 JACOBIAN THEN CONDENSATION WITH REGULARIZATION
C (energy due to viscous regularization is calculated)
SUBROUTINE UMAT(STRESS,STATEV,DDSDDE,SSE,SPD,SCD,
1 RPL,DDSDDT,DRPLDE,DRPLDT,
2
STRAN,DSTRAN,TIME,DTIME,TEMP,DTEMP,PREDEF,DPRED,CMNAME,
3
NDI,NSHR,NTENS,NSTATV,PROPS,NPROPS,COORDS,DROT,PNEWDT,
4
CELENT,DFGRD0,DFGRD1,NOEL,NPT,LAYER,KSPT,KSTEP,KINC)
C
INCLUDE 'ABA_PARAM.INC'
C
CHARACTER*80 CMNAME
DIMENSION STRESS(NTENS),STATEV(NSTATV),
1 DDSDDE(NTENS,NTENS),
2 DDSDDT(NTENS),DRPLDE(NTENS),
3
STRAN(NTENS),DSTRAN(NTENS),TIME(2),PREDEF(1),DPRED(1),
4
PROPS(NPROPS),COORDS(3),DROT(3,3),DFGRD0(3,3),DFGRD1(3,3)

DIMENSION STRANT(6),TSTRANT(4)
DIMENSION CFULL(6,6),CDFULL(6,6)
DIMENSION DDFDE(6), DDMDE(6), DCDDF(6,6), DCDDM(6,6)
DIMENSION ATEMP1(6), ATEMP2(6), TDDSDDE(6,6)
DIMENSION OLD_STRESS(6)
DIMENSION DOLD_STRESS(6),D_STRESS(6)
PARAMETER (ZERO = 0.D0,ONE = 1.D0,TWO = 2.D0, HALF =
0.5D0)
C****************************
C STRANT..... STRAIN AT THE END OF THE INCREMENT
A Apendices 106


C TSTRANT.....TEMPORARY ARRAY TO HOLD THE STRAIN FOR PLANE
STRESS PROBLEM
C CFULL.......FULL 6X6 ELASTICITY MATRIX
C CDFULL......FULL 6X6 DAMAGED ELASTICITY MATRIX
C DDFDE....... D DF/D E
C DDMDE....... D DM/D E
C DCDDF....... D C/ D DF THE DERIVATIVE OF THE FULL MATRIX
OVER DF
C DCDDM........D C/ D DM THE DERIVATIVE OF THE FULL MATRIX
OVER DM
C ATEMP1,ATEMP2...TEMPORARY ARRAY USED IN JACOBIAN
CALCULATION
C TDDSDDE.....UNCONDENSED JACOBIAN MATRIX FOR PLANE STRESS
PROBLEM
C OLD_STRESS...STRESS AT THE BEGINNING OF THE INCREMENT,
SAVED FOR THE ENERGY
C COMPUTATION
C DOLD_STRESS...STRESS AT THE BEGINNING OF THE INCREMENT,
C IF THERE'S NO VISCOUS REGULARIZATION
C D_STRESS...STRESS IF THERE'S NO VISCOUS REGULARIZATION,
THE ABOVE IS CALCULATED
C TO CALCULATE THE SCD, ENERGY CAUSED BY
VISCOUS REGULARIZATION
C STATEV(1) damage variable df
C STATEV(2) damage variable dm
C STATEV(3) regularized damage variable dfv
C STATEV(4) regularizaed damage variable dmv
C STATEV(5:10) TEMPORARY ARRAYS TO SAVE DOLD_STRESS
C************
C
C GET THE MATERIAL PROPERTIES---ENGINEERING CONSTANTS
C
TENL = PROPS(1) !YOUNG'S MODULUS IN DIRECTION
1 (L)
TENT = PROPS(2) !YOUNG'S MODULUS IN DIRECTION
2 (T)
SHRLT = PROPS(3) !SHEAR MODULUS IN 12 PLANE
SHRTT = PROPS(4) !SHEAR MODULUS IN 23 PLANE
A Apendices 107


XNULT = PROPS(5) !POISON'S RATIO POI_12
XNUTT = PROPS(6) !POISON'S RATIO POI_23
XNUTL = XNULT / TENL * TENT !POI_21
C
C GET THE FAILURE PROPERTIES
C
SIGTL = PROPS(7) !FAILURE STRESS IN 1 DIRECTION
IN TENSION
SIGCL = PROPS(8) !FAILURE STRESS IN 1 DIRECTION
IN COMPRESSION
SIGTT = PROPS(9) !FAILURE STRESS IN 2 DIRECTION
IN TENSION
SIGCT = PROPS(10) !FAILURE STRESS IN 2
DIRECTION IN COMPRESSION
SIGSLT = PROPS(11) !FAILURE STRESS IN SHEAR IN 1-
2 PLANE
GFMAT = PROPS(12) !FRACTURE ENERGY IN MATRIX
GFFIB = PROPS(13) !FRACTURE ENERGY IN FIBER
ETA = PROPS(14) ! VISCOSITY FOR REGULARIZATION
C
C CALCULATE THE STRAIN AT THE END OF THE INCREMENT
C
DO I = 1, NTENS
STRANT(I) = STRAN(I) + DSTRAN(I)
END DO
C
C FILL THE 6X6 FULL STIFFNESS MATRIX
DO I = 1, 6
DO J = 1, 6
CFULL(I,J)=ZERO
END DO
END DO
ATEMP = ONE - TWO * XNULT * XNUTL - XNUTT ** TWO
1 - TWO * XNULT * XNUTL * XNUTT
CFULL(1,1) = TENL * (ONE - XNUTT ** TWO) / ATEMP
CFULL(2,2) = TENT * (ONE - XNULT * XNUTL) / ATEMP
CFULL(3,3) = CFULL(2,2)
A Apendices 108


CFULL(1,2) = TENT * (XNULT + XNULT * XNUTT) / ATEMP
CFULL(1,3) = CFULL(1,2)
CFULL(2,3) = TENT * (XNUTT + XNULT * XNUTL) / ATEMP
CFULL(4,4) = SHRLT
CFULL(5,5) = SHRLT
CFULL(6,6) = SHRTT
DO I = 2, 6
DO J = 1, I-1
CFULL(I,J) = CFULL(J,I)
END DO
END DO
C calculate the failure strain by failure stress
EPITL = SIGTL / cfull(1,1) !FAILURE STRAIN 1 DIRECTION
IN TENSION
EPICL = SIGCL / cfull(1,1) !FAILURE STRAIN 1 DIRECTION
IN COMPRESSION
EPITT = SIGTT / cfull(2,2) !TENSILE FAILURE STRAIN 2
DIRECTION
EPICT = SIGCT / cfull(2,2) !COMPRESSIVE FAILURE STRAIN 2
DIRECTION
EPISLT = SIGSLT/ SHRLT ! FAILURE SHEAR STRAIN
...ENGINEERING STRAIN
C
C CHECK THE FAILURE INITIATION CONDITION
C
DFOLD = STATEV(1)
DMOLD = STATEV(2)
DFVOLD = STATEV(3)
DMVOLD = STATEV(4)
CALL
CheckFailureIni(EPITL,EPICL,EPITT,EPICT,EPISLT,STRANT,
1 GFMAT,GFFIB, CELENT, CFULL, DF, DM, DDFDE, DDMDE,
NTENS,
2 DFOLD, DMOLD,NDI)
C
C ! USE VISCOUS REGULARIZATION
C
A Apendices 109


DFV = ETA / (ETA + DTIME) * DFVOLD + DTIME / (ETA +
DTIME) * DF
DMV = ETA / (ETA + DTIME) * DMVOLD + DTIME / (ETA +
DTIME) * DM
C SAVE THE OLD STRESS TO OLD_STRESS
DO I = 1, NTENS
OLD_STRESS(I) = STRESS(I)
END DO

C CALL ROUTINE TO CALCULATE THE STRESS
C CALCULATE THE STRESS IF THERE'S NO VISCOUS
REGULARIZATION
CALL
GetStress(CFULL,CDFULL,DF,DM,D_STRESS,STRANT,NDI,NTENS)
C CALCULATE THE STRESS IF THERE'S VISCOUS REGULARIZATION
CALL
GetStress(CFULL,CDFULL,DFV,DMV,STRESS,STRANT,NDI,NTENS)
C GET THE OLD STRESS IF THERE'S NO VISCOUS REGULARIZATION
DO I=1,NTENS
DOLD_STRESS(I)=STATEV(I+4)
END DO
C SAVE THE CURRENT STRESS IF THERE'S NO VISCOUS
REGULARIZATION
DO I=1,NTENS
STATEV(I+4)=D_STRESS(I)
END DO
C
C CALCULATE THE DERIVATIVE MATRIX DC/DDM, DC/DDF OF THE
DAMAGED MATRIX
C
CALL ElasticDerivative(CFULL,DMV,DFV, DCDDM,DCDDF)
C
C UPDATE THE JACOBIAN
C
C FULL 3D CASE
IF (NDI .EQ. 3) THEN
DO I = 1, NTENS
A Apendices 110


ATEMP1(I) = ZERO
DO J = 1, NTENS
ATEMP1(I) = ATEMP1(I) + DCDDM(I,J) * STRANT(J)
END DO
END DO

DO I = 1, NTENS
ATEMP2(I) = ZERO
DO J = 1, NTENS
ATEMP2(I) = ATEMP2(I) + DCDDF(I,J) * STRANT(J)
END DO
END DO

DO I = 1, NTENS
DO J = 1, NTENS
DDSDDE(I,J)=CDFULL(I,J) + ( ATEMP1(I) *
DDMDE(J)
1 + ATEMP2(I) * DDFDE(J) ) * DTIME / (DTIME
+ ETA)
END DO
END DO
C
C ! PLANE STRESS CASE
C
ELSE IF (NDI .EQ.2) THEN
TSTRANT(1) = STRANT(1)
TSTRANT(2) = STRANT(2)
TSTRANT(3) = -CDFULL(1,3) / CDFULL(3,3) * STRANT(1)
1 - CDFULL(2,3) / CDFULL(3,3) * STRANT(2)
TSTRANT(4) = STRANT(3)
DO I = 1, 4
ATEMP1(I) = ZERO
DO J = 1, 4
ATEMP1(I) = ATEMP1(I) + DCDDM(I,J) * TSTRANT(J)
END DO
END DO
A Apendices 111



DO I = 1, 4
ATEMP2(I) = ZERO
DO J = 1, 4
ATEMP2(I) = ATEMP2(I) + DCDDF(I,J) * TSTRANT(J)
END DO
END DO
DO I = 1,6
DO J = 1,6
TDDSDDE(I,J) = ZERO
END DO
END DO
C TO GET THE UNCONDENSED JACOBIAN FOR PLANE STRESS CASE
DO I = 1, NTENS
DO J = 1, NTENS
DDSDDE(I,J) = ZERO
END DO
END DO
DO I = 1, 4
DO J = 1, 4
TDDSDDE(I,J)=CDFULL(I,J) + ( ATEMP1(I) *
DDMDE(J)
1 + ATEMP2(I) * DDFDE(J) ) * DTIME / (DTIME
+ ETA)
END DO
END DO
C
C CONDENSE THE JACOBIAN MATRIX FOR PLANE STRESS PROBLEM
C
CALL MatrixCondense(TDDSDDE,DDSDDE)
END IF
C
C TO UPDATE THE STATE VARIABLE
C
STATEV(1) = DF
STATEV(2) = DM
A Apendices 112


STATEV(3) = DFV
STATEV(4) = DMV
C
C TO COMPUTE THE ENERGY
C
DO I = 1, NDI
SSE = SSE + HALF * (STRESS(I) + OLD_STRESS(I)) *
DSTRAN(I)
END DO
DO I = NDI+1, NTENS
SSE = SSE + (STRESS(I) + OLD_STRESS(I)) * DSTRAN(I)
END DO
C TO COMPUTE THE INTERNAL ENERGY WITHOUT VISCOUS
REGULARIZATION
DO I = 1, NDI
SCD = SCD + HALF * (STRESS(I) + OLD_STRESS(I)
1 -D_STRESS(I)-DOLD_STRESS(I)) * DSTRAN(I)
END DO
DO I = NDI+1, NTENS
SCD = SCD + (STRESS(I) + OLD_STRESS(I)
1 -D_STRESS(I)-DOLD_STRESS(I)) * DSTRAN(I)
END DO

RETURN
END

C*************************************************************
*****************
C CALCULATE THE STRESS BASED ON THE DAMAGE
VARAIBLES***************************
C*************************************************************
*****************
SUBROUTINE
GetStress(CFULL,CDFULL,DFV,DMV,STRESS,STRANT,NDI,NTENS)
INCLUDE 'ABA_PARAM.INC'
DIMENSION CFULL(6,6),CDFULL(6,6),STRESS(NTENS),
1 STRANT(6),CDTHREE(3,3)
A Apendices 113


PARAMETER (ZERO=0.D0, ONE=1.D0)
C CDTHREE.....DAMAGED CONDENSED-ELASTICITY MATRIX FOR
PLANE STRESS PROBLEM
DO I = 1, 6
DO J = 1, 6
CDFULL(I,J)=CFULL(I,J)
END DO
END DO
IF ( (DFV .NE. ZERO) .OR. (DMV .NE. ZERO)) THEN
CDFULL(1,1) = (ONE - DFV) * CFULL(1,1)
CDFULL(1,2) = (ONE - DFV) * (ONE - DMV) * CFULL(1,2)
CDFULL(2,1) = CDFULL(1,2)
CDFULL(2,2) = (ONE - DMV) * CFULL(2,2)
CDFULL(1,3) = (ONE - DFV) * CFULL(1,3)
CDFULL(3,1) = CDFULL(1,3)
CDFULL(2,3) = (ONE- DMV) * CFULL(2,3)
CDFULL(3,2) = CDFULL(2,3)
CDFULL(4,4) = (ONE - DMV) * (ONE - DFV) * CFULL(4,4)
END IF
C UPDATE THE STRESS STATE IF 3D CASE
C
IF (NDI .EQ. 3) THEN
DO I = 1, NTENS
STRESS(I)=ZERO
DO J = 1, NTENS
STRESS(I)=STRESS(I)+CDFULL(I,J) * STRANT(J)
END DO
END DO

C
C INITIALIZE THE 3X3 CONDENSED STIFFNESS MATRIX IF PLANE
STRESS CASE
C
ELSE IF ( NDI .EQ. 2) THEN
DO I = 1, NTENS
DO J = 1, NTENS
A Apendices 114


CDTHREE(I,J)=ZERO
END DO
END DO
C
C
C CONDENSE THE UNDAMAGED STIFFNESS MATRIX
C
CALL MatrixCondense(CDFULL,CDTHREE)
C
C UPDATE THE STRESS
C
DO I = 1, NTENS
STRESS(I)=ZERO
DO J = 1, NTENS
STRESS(I)=STRESS(I)+CDTHREE(I,J) * STRANT(J)
END DO
END DO
END IF
RETURN
END
C*************************************************************
*****************
C TO CHECK THE FAILURE INITIATION AND THE CORRESPONDING
DERIVATIVE*********
C*************************************************************
*****************
SUBROUTINE
CheckFailureIni(EPITL,EPICL,EPITT,EPICT,EPISLT,STRANT,
1 GFMAT,GFFIB, CELENT, CFULL, DF, DM, DDFDE, DDMDE,
NTENS,
2 DFOLD,DMOLD, NDI )
INCLUDE 'ABA_PARAM.INC'
DIMENSION DDFDE(6), DDMDE(6), STRANT(6), CFULL(6,6)
DIMENSION DFMNDE(6), DFFNDE(6)
PARAMETER (ZERO = 0.D0, ONE = 1.D0, TWO = 2.D0, HALF =
0.5D0)
C
A Apendices 115


C CHECK THE INITIATION CONDITION FOR MATRIX
C FMN=FM/EPITT > 1 THEN EVALUATE THE DAMAGE VARIABLE AND
DERIVATIVE
C
TERM1 = STRANT(2)**TWO / EPICT / EPITT
TERM2 = (EPICT - EPITT) / EPICT / EPITT * STRANT(2)
IF (NDI .EQ. 3) THEN
TERM3 = (STRANT(4))**TWO / EPISLT**TWO
ELSE IF (NDI .EQ. 2) THEN
TERM3 = (STRANT(3))**TWO / EPISLT**TWO
END IF
TERM = TERM1 + TERM2 + TERM3
IF (TERM .GT. ZERO) THEN
FMN = SQRT(TERM)
ELSE
FMN = ZERO
END IF
C
C INITIALIZE THE ARRAY AND VARIABLE
C
DM = ZERO
DDMDFMN = ZERO
DO I = 1, 6
DFMNDE(I) = ZERO
DDMDE(I) = ZERO
END DO
IF (FMN .GT. ONE) THEN
C CALCULATE DM, DDMDFMN
CALL DamageEvaluation( CFULL(2,2), FMN, GFMAT,
CELENT,
1 EPITT, DM, DDMDFMN)
C CALCULATE DFMNDE
IF (DM .GT. DMOLD) THEN
DFMNDE(2) = HALF / FMN * (TWO * STRANT(2) + EPICT
- EPITT)
1 / EPICT / EPITT
A Apendices 116


IF (NDI .EQ. 3) THEN
DFMNDE(4) = ONE / FMN * STRANT(4) / EPISLT**TWO
ELSE IF (NDI .EQ. 2) THEN
DFMNDE(4) = ONE / FMN * STRANT(3) / EPISLT**TWO
END IF
DO I = 1, 6
DDMDE(I) = DFMNDE(I) * DDMDFMN
END DO
END IF
END IF
DM = MAX (DM, DMOLD)
C
C CHECK THE INITIATION CONDITION FOR FIBER
C FFN=FF/EPITL>1 THEN CALCULATE THE DAMAGE VARIABLE AND
DERIVATIVE
C
TERM1 = STRANT(1)**TWO / EPICL / EPITL
TERM2 = (EPICL - EPITL) / EPICL / EPITL *STRANT(1)
TERM = TERM1 + TERM2
IF (TERM .GT. ZERO) THEN
FFN = SQRT(TERM)
ELSE
FFN = ZERO
END IF
DF = ZERO
DDFDFFN = ZERO
DO I = 1, 6
DFFNDE(I) = ZERO
DDFDE(I) = ZERO
END DO
IF (FFN .GT. ONE) THEN
C CALCULATE DF, DDFDFFN
CALL DamageEvaluation( CFULL(1,1), FFN, GFFIB,
CELENT,
1 EPITL, DF, DDFDFFN)
C CALCULATE DFFNDE
A Apendices 117


IF (DF .GT. DFOLD) THEN
DFFNDE(1) = HALF / FFN * (TWO * STRANT(1) +
EPICL - EPITL)
1 / EPICL / EPITL
DDFDE(1) = DFFNDE(1) * DDFDFFN
END IF
END IF
DF = MAX (DF, DFOLD)
RETURN
END
C*************************************************************
*****************
C SUBROUTINE TO EVALUATE THE DAMAGE AND THE
c DERIVATIVE************************
C*************************************************************
*****************
SUBROUTINE DamageEvaluation(STIFF, FN, GF, CELENT, EPIT,
D,
1 DDDFN)
C CALCULATE DAMAGE VARIABLE
INCLUDE 'ABA_PARAM.INC'
PARAMETER (ONE = 1.D0,tol=1d-3, zero = 0.d0)
TERM1 = STIFF * EPIT**2 * CELENT / GF
TERM2 = (ONE - FN) * TERM1
D = ONE - EXP(TERM2) / FN
C CALCULATE THE DERIVATIVE OF DAMAGE VARIABLE WITH RESPECT
TO FAILURE
C RITERION
DDDFN = (ONE / FN + TERM1) * (ONE - D)
RETURN
END
C*************************************************************
*****************
C SUBROUTINE TO CONDENSE THE 4X4 MATRIX INTO 3X3
MATRIX********************
C*************************************************************
*****************
SUBROUTINE MatrixCondense(CFULL,CTHREE)
A Apendices 118


INCLUDE 'ABA_PARAM.INC'
DIMENSION CFULL(6,6),CTHREE(3,3)
C
CTHREE(1,1) = CFULL(1,1) - CFULL(1,3) * CFULL(3,1) /
CFULL(3,3)
CTHREE(1,2) = CFULL(1,2) - CFULL(1,3) * CFULL(3,2) /
CFULL(3,3)
CTHREE(2,1) = CFULL(2,1) - CFULL(2,3) * CFULL(3,1) /
CFULL(3,3)
CTHREE(2,2) = CFULL(2,2) - CFULL(2,3) * CFULL(3,2) /
CFULL(3,3)
CTHREE(3,3) = CFULL(4,4)
RETURN
END
C*************************************************************
******************
C SUBROUTINE TO GET THE DERIVATIVE MATRIX OF CONDENSE
DAMAGED MATRIX OVER
C**** THE DAMAGE
VARIABLE******************************************************
C*************************************************************
******************
SUBROUTINE ElasticDerivative(CFULL,DMV,DFV, DCDDM,DCDDF)
INCLUDE 'ABA_PARAM.INC'
DIMENSION CFULL(6,6), DCDDM(6,6),
1 DCDDF(6,6)
PARAMETER (ZERO = 0.D0, ONE = 1.D0, HALF = 0.5D0)
C initialize the data to zero
DO I = 1, 6
DO J = 1, 6
DCDDM(I,J) = ZERO
DCDDF(I,J) = ZERO
END DO
END DO
C
C CALCULATE DC/DDF
C
DCDDF(1,1) = - CFULL(1,1)
A Apendices 119


DCDDF(1,2) = - (ONE - DMV) * CFULL(1,2)
DCDDF(2,1) = DCDDF(1,2)
DCDDF(1,3) = -CFULL(1,3)
DCDDF(3,1) = DCDDF(1,3)
DCDDF(4,4) = -(ONE - DMV) * CFULL(4,4)
C
C CALCULATE DC/DDM
C
DCDDM(1,2) = - (ONE - DFV) * CFULL(1,2)
DCDDM(2,1) = DCDDM(1,2)
DCDDM(2,2) = -CFULL(2,2)
DCDDM(2,3) = -CFULL(2,3)
DCDDM(3,2) = DCDDM(2,3)
DCDDM(4,4) = -(ONE - DFV) * CFULL(4,4)
RETURN
END

A Apendices 120



A Apendices 121


b. List of Figures
Fig. 2.1: Stress-strain curves for typical fibers used in reinforced composite ............................ 6
Fig. 2.2: Fiber processability tests [5]: (a) straight tow, (b) tow with a loop and (c) tow with a
knot ............................................................................................................................................. 7
Fig. 2.3: The levels of analysis for a structure made of laminated composite. [7] ..................... 9
Fig. 2.4: Composite material systems. [7] ................................................................................. 9
Fig. 2.5: Illustration of the matrix, fiber, and void volumes. [7] ............................................. 10
Fig. 2.6: Representative elements: Element 1 (left) and Element 2 (right). [7] ...................... 11
Fig. 2.7: Element 1 subjected to a force in the fiber direction (left), and Element 2 subjected
to a force in the transverse direction (right). [7] ....................................................................... 13
Fig. 2.8: Element 2 subjected to a shear force (left) and deformation of the top (ijkl) surface
(right). [7] ................................................................................................................................. 15
Fig. 2.9: Deformation of Element 1 subjected to an axial force in the x
1
direction. [7] ........... 17
Fig. 2.10: Anisotropic fiber orientation for plane-strain condition. ........................................ 19
Fig. 2.11: Monoclinic fiber orientation for plane-strain condition. ......................................... 20
Fig. 2.12: Orthotropic fiber orientation for plane-strain condition.......................................... 21
Fig. 2.13: Transversely Isotropic fiber orientation for plane-strain condition. ....................... 22
Fig. 2.14: Isotropic fiber orientation for plane-strain condition. ............................................. 23
Fig. 2.15: Coordinate Base Transformation ............................................................................ 25
Fig. 2.16: Laminate under plane-stress conditions (left) and deformation of the plane in the x-
z plane (right). .......................................................................................................................... 28
Fig. 2.17: Modes of failure for bolted joints in advanced composites. ................................... 40
Fig. 2.18: Schematics of delaminations caused by drilling (a) upon entry, and (b) upon exit.
[16] ........................................................................................................................................... 43
Fig. 2.19: Typical failure modes of composites. ..................................................................... 30
Fig. 2.20: Failure surfaces for some failure criteria. [17] ........................................................ 34
Fig. 2.21: Double Cantilever Beam undergoing crack growth. [32] ........................................ 48
Fig. 2.22: Fracture mechanics crack propagation modes. ........................................................ 48
Fig. 2.23: VCCT approach for pure Mode I: Crack Extension (left), Crack Growth Criterion
(right). [22] ............................................................................................................................... 51
Fig. 2.24: Cohesive Zone in a double cantilever beam. [34] .................................................... 51
Fig. 2.25: Cohesive Zone ahead of delamination tip. [34] ....................................................... 52
Fig. 2.26: Constitutive strain softening equations. [34] ........................................................... 53
A Apendices 122


Fig. 2.27: Linear elastic-linear constitutive equation. [34]....................................................... 54
Fig. 2.28: Mixed-mode behavior for linear elastic-linear constitutive equation. [34] .............. 55
Fig. 3.1: Drawing of the Composite Coupon .......................................................................... 58
Fig. 3.2: Drawing of the Protruding-Head Bolt ....................................................................... 59
Fig. 3.3: Drawing of the Test Fixture ...................................................................................... 59
Fig. 3.4: Drawing of the Retaining Block ............................................................................... 60
Fig. 3.5: Drawing of the Countersunk-Head Bolt ................................................................... 60
Fig. 3.6: Drawing of the Semi-Circumferential Part ............................................................... 61
Fig. 3.7: ARCAN Test under 45 loading (left) and under 90 loading (right) ........................ 61
Fig. 3.8: User-Interface Test Data Selection ............................................................................ 63
Fig. 3.9: User-Interface Coupon Data Selection....................................................................... 64
Fig. 3.10: User-Interface Pretension/Displacement Selection .................................................. 64
Fig. 3.11: User-Interface Abaqus Version/Input File ............................................................... 65
Fig. 3.12: Nut (left) and Protruding-Head Bolt (right) meshed ................................................ 66
Fig. 3.13: Retaining Block meshed ......................................................................................... 67
Fig. 3.14: Countersunk-Head Bolt (left) and Nut (right) meshed ........................................... 67
Fig. 3.15: Core Vicinity meshed .............................................................................................. 69
Fig. 3.16: Coupon Core meshed .............................................................................................. 70
Fig. 3.17: Coupon Outer Core meshed .................................................................................... 71
Fig. 3.18: Spatial representation of a three-dimensional cohesive element (left) and
deformation modes of a cohesive element (right). [21]............................................................ 73
Fig. 3.19: Topology of an eroding contact surface. [21] ......................................................... 76
Fig. 3.20: Retaining Fastener Connector ................................................................................. 77
Fig. 3.21: Retaining Fastener Connector ................................................................................. 81
Fig. 4.1: Pure Tension Experimental Result with Protruding Head Bolt ................................ 83
Fig. 4.2: Experimental Result of Pure Tension Test................................................................ 84
Fig. 4.3: Pure Shear Experimental Result with Protruding Head Bolt .................................... 84
Fig. 4.4: Experimental Result of Pure Shear Test ................................................................... 85
Fig. 4.5: Comparison of the Experimental Results in Terms of Loading Angle for Protruding
Head Bolt .................................................................................................................................. 85
Fig. 4.6: Comparison of the Experimental Results in Terms of Loading Angle for
Countersunk Head Bolt ............................................................................................................ 87
A Apendices 123


Fig. 4.7: Comparison of the Experimental Results in Terms of Shear-Tension Maximum
Load Peak ................................................................................................................................. 88
Fig. 5.1: Energy Plot for Simulation without Cohesives .......................................................... 91
Fig. 5.2: Simulation without Cohesives ................................................................................... 92
Fig. 5.3: Deformed and Un-deformed Plot of the Central Core for Model without Cohesives
(Pure Tension) .......................................................................................................................... 93
Fig. 5.4: Energy Plot for Simulation without Cohesives .......................................................... 94
Fig. 5.5: Simulation without Cohesives .................................................................................. 94
Fig. 5.6: Deformed and Un-deformed Plot of the Central Core for Model without Cohesives
(Pure Shear) .............................................................................................................................. 95
Fig. 5.7: Simulation with Cohesive Elements .......................................................................... 96
Fig. 5.8: Simulation with Cohesive Elements .......................................................................... 96
Fig. 5.9: Deformed and Un-deformed Plot of the Central Core for Cohesive Elements Model
(Pure Tension) .......................................................................................................................... 97
Fig. 5.10: Simulation with Cohesive Elements ........................................................................ 98
Fig. 5.11: Deformed and Un-deformed Plot of the Central Core for Cohesive Contact Model
(Pure Tension) .......................................................................................................................... 99
Fig. 6.1: Evolution of the computational time with the number of CPUs for the simulation
with the approach of Cohesive Elements ............................................................................... 101


A Apendices 124




A Apendices 125


c. List of Tables
Tab. 1: Evolution of final energy consumption in transport, by transport mode, 1990-2004,
EU-25 (in million toe) [2] 2
Tab. 2: Functional flowchart between the user and the tool. 3
Tab. 3: Fiber and Wire Properties [3] 5
Tab. 4: Types of Matrix 8
Tab. 5: Tasks distribution between scripts 63
Tab. 6: Applied Element Types 68
Tab. 7: Meshing and Element Controls of Central Nut 68
Tab. 8: Meshing and Element Controls of Central Bolt PAN-Head 69
Tab. 9: Meshing and Element Controls of the Core Vicinity 69
Tab. 10: Meshing and Element Controls of the Coupon Core 70
Tab. 11: Meshing and Element Controls of the Coupon Outer Cores 71
Tab. 12: Meshing and Element Controls of Retaining Block 72
Tab. 13: Meshing and Element Controls of Retaining Block 72
Tab. 14: Meshing and Element Controls of the Retaining Bolts 73
Tab. 15: Meshing and Element Controls of the Retaining Nuts 73
Tab. 16: Shear-Tension Failure Modes for Protruding Head Bolt 86
Tab. 17: Shear-Tension Failure Modes for Countersunk Head Bolt 88
Tab. 18: Comparison of the Approaches in Terms of Computational Cost 101


A Apendices 126


B Bibliography 127


B. Bibliography
[1] T. Richardson. Composites: A Design Guide. Industrial Press Inc., New York, 1987.
[2] Eurostat. European Comision. http://epp.eurostat.ec.europa.eu/statistics_explained/
index.php/Transport_energy_consumption_and_emissions.
[3] Albert G. H. Dietz, Composite Materials, 1965 Edgar Marburg Lecture, American
Society for Testing and Materials, 1965.
[4] E.A. Rothman and G.E. Molten. STP, February 1969, ASTM, Philadelphia, Pa, USA.
[5] V. V. Vasiliev and E. V. Morozov. Mechanics and analysis of composite materials.
Elsevier science Ltd, 2001.
[6] T. Richardson. Composites: A Design Guide. Industrial Press Inc., New York, 1987.
[7] L. P. Kollar and G. S. Springer. Mechanics of composite structures. Cambridge
University Press, 2003.
[8] Sanjay K. Mazumdar. Composites Manufacturing, materials, product and process
engineering. CRC Press LLC, 2002.
[9] J. M. Whitney, D. L. Stansbarger and H. B. Howell. J. Comp: Mterials, January 1971, 5,
24.
[10] T. A. Collings. Experimentally Determined Strength of Mechanically Fastened Joints.
[11] T.A. Collings. The Strength of Bolted Joints in Multi-Directional CFRP Laminates.
RAE Technical Report 75127, 1975.
[12] A. Baker, S.Dutton and D.Kelly. Composite Materials for Aircraft Structures. AIAA
Education Serie, second edition, 2004.
[13] H. - J. Park. Effects of stacking sequence and clamping force on the bearing strengths of
the mechanically fastened joints in composite laminates. Composite Structures, 53:213-
221, August 2001.
[14] C. Poon. Literature Review on the Design of Mechanically Fastened Composite Joints.
[15] H. M. Wittmeyer and F. E. Smode. "Fatigue Resistant Fastener," in Aircraft Structures
and Materials Application, Proceedings of the Ist National SAMPE Technical
Conference, Seattle, Washington, September 9-11, 1969.
[16] Persson et al., Composites, 28A, 141, 1997. Elsevier Science Ltd., U.K.
[17] C. T. Sun, B. J. Quinn, J. Tao and D. W. Oplinger. Comparative evaluation of failure
analysis methods for composite materials. Technical Report DOT/ FAA/ AR-95/109,
Office of Aviation Research, Washington, D. C. 20591, 1996.
[18] G. M. K. Pearce. High strain-rate behavior of bolted joints in carbon fibres composite
structures. PhD thesis, School of Mechanical and Manufacturing Engineering,
University of New Soth Wales, Sydney, October 2009.
B Bibliography 128


[19] W. H. Chen, S. S. Lee and J. T. Yeh. Three-dimentional contact stress analysis of a
composite laminate with bolted joint. Composite structures, 30:281-297, 1995.
[20] M. McCarthy. Bojcas: Bolted joints in composite aircraft structures. Air & Space
Europe, 3:139-142, August 2001.
[21] Abaqus 6.11 Documentation
[22] S. Drapier. Dynamique des solides et des structures. Saint-Etienne, France, October
2005.
[23] T. Belytschko, W. K. Liu and B. Moran. Non Linear Finite Elements for Continua and
Structures. Wiley, first edition, 2000.
[24] J. Mackerle. Finite element analysis of fastening and joining: A bibliography (1990-
2002). International Journal of Pressure Vessels and Piping, 80:253-271, February
2003.
[25] M. L. Dano, E. Kamal and G. Gendron. Analysis of bolted joints in composite
laminates-strains and bearing stiffness prediction. Composite structures, 79:562-570,
2006.
[26] M. A. McCarthy, C. T. McCarthy, V. P. Lawlor and W. F. Stanley. Three-dimensional
finite element analysis of single-bolt, single-lap composite bolted joints: part i- model
development and validation. Composite structures, 71:140-158, November 2005.
[27] J. Ekh and J. Schon. Effect of secondary bending on strength prediction of composite,
single shear lap joints. Composite science and technology, 65:953-965, December 2004.
[28] S. Kapti, O. Sayman and S. Benli. Experimental and numerical failure analysis of
carbon/epoxy laminated composite joints under different conditions. Materials and
Design, 31:4933-4942, May 2010.
[29] M. A. McCarthy, C. T. McCarthy, V. P. Lawlor and W. F. Stanley. Progressive analysis
of multi-bolt composite joints with variable bolt-hole claerances. Composit: Part B,
36:290-305, January 2005.
[30] V. N. Burlayenkol and T. Sadowski. FE modeling of delamination growth in
interlaminar fracture specimens. Budownictwo I Architektura, 2:95-109, 2008.
[31] G. Wimmer, C. Schuecker and H. E. Pettermann. Numerical simulation of delamination
in laminated composite componets-a combination of a strength criterion and fracture
mechanics. Composites: Part B, 40:158-165, January 2009.
[32] T.Dieh. Modeling surface-bonded structures with abaqus cohesive elements: Beam-type
solutions. In 2005 Abaqus Users Conference, 2005.
[33] P. P. Camahno, C. G. Davila and D.R. Ambur. Numerical simulation of delamination
growth in material composites. Technical Report NASA/TP-2001-211041, National
Aeronautics and Space Administration, Langley Research Center, Hampton, Virginia
23681-2199, 2011.
B Bibliography 129


[34] P. W. Harper and S. R. Hallet. Cohesive zone length in numerical simulation of
composite delamination. Engineering Fracture Mechanics, 75:4774-4792, November
2008.
[35] S. T. Pinho, L. Iannucci and P. Robinson. Formulation and implementation of the de-
cohesion elements in an explicit finite element code. Composites: Part A, 37:778-789,
May 2006.
[36] P. Linde, J. Pleitner, H. de Boer and C. Carmone. Modeling and simulation of fiber
metal laminates. In 2004 Abaqus Users Conference, 2004.

Das könnte Ihnen auch gefallen